Merge branch 'develop' into madlittlemods/11850-migrate-to-opentelemetry

Conflicts:
	synapse/handlers/message.py
	synapse/logging/opentracing.py

148 files changed, 4277 insertions, 1653 deletions

diff --git a/.github/dependabot.yml b/.github/dependabot.yml
new file mode 100644
index 0000000000..9c7db1fc86
--- /dev/null
+++ b/.github/dependabot.yml
@@ -0,0 +1,17 @@
+version: 2
+updates:
+  - # "pip" is the correct setting for poetry, per https://docs.github.com/en/code-security/dependabot/dependabot-version-updates/configuration-options-for-the-dependabot.yml-file#package-ecosystem
+    package-ecosystem: "pip"
+    directory: "/"
+    schedule:
+      interval: "weekly"
+
+  - package-ecosystem: "docker"
+    directory: "/docker"
+    schedule:
+      interval: "weekly"
+
+  - package-ecosystem: "github-actions"
+    directory: "/"
+    schedule:
+      interval: "weekly"
diff --git a/CHANGES.md b/CHANGES.md
index 82b5526f94..fbb57f0e04 100644
--- a/CHANGES.md
+++ b/CHANGES.md
@@ -1,17 +1,25 @@
-Synapse 1.68.0rc2 (2022-09-23)
-==============================
+Synapse 1.68.0 (2022-09-27)
+===========================
 
-Please note that Synapse will now refuse to start if configured to use a version of SQLite earlier than 3.27.
+Please note that Synapse will now refuse to start if configured to use a version of SQLite older than 3.27.
 
 In addition, please note that installing Synapse from a source checkout now requires a recent Rust compiler.
 Those using packages will not be affected. On most platforms, installing with `pip install matrix-synapse` will not be affected.
-See the [upgrade notes](https://matrix-org.github.io/synapse/v1.68/upgrade.html#upgrading-to-v1670).
+See the [upgrade notes](https://matrix-org.github.io/synapse/v1.68/upgrade.html#upgrading-to-v1680).
+
+Bugfixes
+--------
+
+- Fix packaging to include `Cargo.lock` in `sdist`. ([\#13909](https://github.com/matrix-org/synapse/issues/13909))
+
 
+Synapse 1.68.0rc2 (2022-09-23)
+==============================
 
 Bugfixes
 --------
 
-- Fix building from packaged sdist. Broke in v1.68.0rc1. ([\#13866](https://github.com/matrix-org/synapse/issues/13866))
+- Fix building from packaged sdist. Broken in v1.68.0rc1. ([\#13866](https://github.com/matrix-org/synapse/issues/13866))
 
 
 Internal Changes
@@ -19,7 +27,7 @@ Internal Changes
 
 - Fix the release script not publishing binary wheels. ([\#13850](https://github.com/matrix-org/synapse/issues/13850))
 - Lower minimum supported rustc version to 1.58.1. ([\#13857](https://github.com/matrix-org/synapse/issues/13857))
-- Lock Rust dependencies versions. ([\#13858](https://github.com/matrix-org/synapse/issues/13858))
+- Lock Rust dependencies' versions. ([\#13858](https://github.com/matrix-org/synapse/issues/13858))
 
 
 Synapse 1.68.0rc1 (2022-09-20)
@@ -40,7 +48,7 @@ Features
 Bugfixes
 --------
 
-- Fix a bug introduced in Synapse v1.41.0 where the `/hierarchy` API returned non-standard information (a `room_id` field under each entry in `children_state`). ([\#13506](https://github.com/matrix-org/synapse/issues/13506))
+- Fix a bug introduced in Synapse 1.41.0 where the `/hierarchy` API returned non-standard information (a `room_id` field under each entry in `children_state`). ([\#13506](https://github.com/matrix-org/synapse/issues/13506))
 - Fix a long-standing bug where previously rejected events could end up in room state because they pass auth checks given the current state of the room. ([\#13723](https://github.com/matrix-org/synapse/issues/13723))
 - Fix a long-standing bug where Synapse fails to start if a signing key file contains an empty line. ([\#13738](https://github.com/matrix-org/synapse/issues/13738))
 - Fix a long-standing bug where Synapse would fail to handle malformed user IDs or room aliases gracefully in certain cases. ([\#13746](https://github.com/matrix-org/synapse/issues/13746))
@@ -54,10 +62,10 @@ Improved Documentation
 ----------------------
 
 - Note that `libpq` is required on ARM-based Macs. ([\#13480](https://github.com/matrix-org/synapse/issues/13480))
-- Fix a mistake in the config manual: the `event_cache_size` _is_ scaled by `caches.global_factor`. The documentation was incorrect since Synapse v1.22.0. ([\#13726](https://github.com/matrix-org/synapse/issues/13726))
+- Fix a mistake in the config manual introduced in Synapse 1.22.0: the `event_cache_size` _is_ scaled by `caches.global_factor`. ([\#13726](https://github.com/matrix-org/synapse/issues/13726))
 - Fix a typo in the documentation for the login ratelimiting configuration. ([\#13727](https://github.com/matrix-org/synapse/issues/13727))
 - Define Synapse's compatability policy for SQLite versions. ([\#13728](https://github.com/matrix-org/synapse/issues/13728))
-- Add docs for common fix of deleting the `matrix_synapse.egg-info/` directory for fixing Python dependency problems. ([\#13785](https://github.com/matrix-org/synapse/issues/13785))
+- Add docs for the common fix of deleting the `matrix_synapse.egg-info/` directory for fixing Python dependency problems. ([\#13785](https://github.com/matrix-org/synapse/issues/13785))
 - Update request log format documentation to mention the format used when the authenticated user is controlling another user. ([\#13794](https://github.com/matrix-org/synapse/issues/13794))
 
 
@@ -98,7 +106,7 @@ Internal Changes
 - Fix a memory leak when running the unit tests. ([\#13798](https://github.com/matrix-org/synapse/issues/13798))
 - Use partial indices on SQLite. ([\#13802](https://github.com/matrix-org/synapse/issues/13802))
 - Check that portdb generates the same postgres schema as that in the source tree. ([\#13808](https://github.com/matrix-org/synapse/issues/13808))
-- Fix Docker build when Rust .so has been build locally first. ([\#13811](https://github.com/matrix-org/synapse/issues/13811))
+- Fix Docker build when Rust .so has been built locally first. ([\#13811](https://github.com/matrix-org/synapse/issues/13811))
 - Complement: Initialise the Postgres database directly inside the target image instead of the base Postgres image to fix building using Buildah. ([\#13819](https://github.com/matrix-org/synapse/issues/13819))
 - Support providing an index predicate clause when doing upserts. ([\#13822](https://github.com/matrix-org/synapse/issues/13822))
 - Minor speedups to linting in CI. ([\#13827](https://github.com/matrix-org/synapse/issues/13827))
@@ -152,7 +160,7 @@ Bugfixes
 - Fix [MSC3030](https://github.com/matrix-org/matrix-spec-proposals/pull/3030) `/timestamp_to_event` endpoint to return the correct next event when the events have the same timestamp. ([\#13658](https://github.com/matrix-org/synapse/issues/13658))
 - Fix bug where we wedge media plugins if clients disconnect early. Introduced in v1.22.0. ([\#13660](https://github.com/matrix-org/synapse/issues/13660))
 - Fix a long-standing bug which meant that keys for unwhitelisted servers were not returned by `/_matrix/key/v2/query`. ([\#13683](https://github.com/matrix-org/synapse/issues/13683))
-- Fix a bug introduced in Synapse v1.20.0 that would cause the unstable unread counts from [MSC2654](https://github.com/matrix-org/matrix-spec-proposals/pull/2654) to be calculated even if the feature is disabled. ([\#13694](https://github.com/matrix-org/synapse/issues/13694))
+- Fix a bug introduced in Synapse 1.20.0 that would cause the unstable unread counts from [MSC2654](https://github.com/matrix-org/matrix-spec-proposals/pull/2654) to be calculated even if the feature is disabled. ([\#13694](https://github.com/matrix-org/synapse/issues/13694))
 
 
 Updates to the Docker image
@@ -179,7 +187,7 @@ Deprecations and Removals
 
 - Drop support for calling `/_matrix/client/v3/rooms/{roomId}/invite` without an `id_access_token`, which was not permitted by the spec. Contributed by @Vetchu. ([\#13241](https://github.com/matrix-org/synapse/issues/13241))
 - Remove redundant `_get_joined_users_from_context` cache. Contributed by Nick @ Beeper (@fizzadar). ([\#13569](https://github.com/matrix-org/synapse/issues/13569))
-- Remove the ability to use direct TCP replication with workers. Direct TCP replication was deprecated in Synapse v1.18.0. Workers now require using Redis. ([\#13647](https://github.com/matrix-org/synapse/issues/13647))
+- Remove the ability to use direct TCP replication with workers. Direct TCP replication was deprecated in Synapse 1.18.0. Workers now require using Redis. ([\#13647](https://github.com/matrix-org/synapse/issues/13647))
 - Remove support for unstable [private read receipts](https://github.com/matrix-org/matrix-spec-proposals/pull/2285). ([\#13653](https://github.com/matrix-org/synapse/issues/13653), [\#13692](https://github.com/matrix-org/synapse/issues/13692))
 
 
@@ -223,7 +231,7 @@ was originally planned for Synapse 1.64, but was later deferred until now. See
 the [upgrade notes](https://matrix-org.github.io/synapse/v1.66/upgrade.html#upgrading-to-v1660) for more details.
 
 Deployments with multiple workers should note that the direct TCP replication
-configuration was deprecated in Synapse v1.18.0 and will be removed in Synapse
+configuration was deprecated in Synapse 1.18.0 and will be removed in Synapse
 v1.67.0. In particular, the TCP `replication` [listener](https://matrix-org.github.io/synapse/v1.66/usage/configuration/config_documentation.html#listeners)
 type (not to be confused with the `replication` resource on the `http` listener
 type) and the `worker_replication_port` config option will be removed .
@@ -353,7 +361,7 @@ Bugfixes
 --------
 
 - Update the version of the LDAP3 auth provider module included in the `matrixdotorg/synapse` DockerHub images and the Debian packages hosted on packages.matrix.org to 0.2.2. This version fixes a regression in the module. ([\#13470](https://github.com/matrix-org/synapse/issues/13470))
-- Fix a bug introduced in Synapse v1.41.0 where the `/hierarchy` API returned non-standard information (a `room_id` field under each entry in `children_state`) (this was reverted in v1.65.0rc2, see changelog notes above). ([\#13365](https://github.com/matrix-org/synapse/issues/13365))
+- Fix a bug introduced in Synapse 1.41.0 where the `/hierarchy` API returned non-standard information (a `room_id` field under each entry in `children_state`) (this was reverted in v1.65.0rc2, see changelog notes above). ([\#13365](https://github.com/matrix-org/synapse/issues/13365))
 - Fix a bug introduced in Synapse 0.24.0 that would respond with the wrong error status code to `/joined_members` requests when the requester is not a current member of the room. Contributed by @andrewdoh. ([\#13374](https://github.com/matrix-org/synapse/issues/13374))
 - Fix bug in handling of typing events for appservices. Contributed by Nick @ Beeper (@fizzadar). ([\#13392](https://github.com/matrix-org/synapse/issues/13392))
 - Fix a bug introduced in Synapse 1.57.0 where rooms listed in `exclude_rooms_from_sync` in the configuration file would not be properly excluded from incremental syncs. ([\#13408](https://github.com/matrix-org/synapse/issues/13408))
@@ -418,7 +426,7 @@ No significant changes since 1.64.0rc2.
 Deprecation Warning
 -------------------
 
-Synapse v1.66.0 will remove the ability to delegate the tasks of verifying email address ownership, and password reset confirmation, to an identity server.
+Synapse 1.66.0 will remove the ability to delegate the tasks of verifying email address ownership, and password reset confirmation, to an identity server.
 
 If you require your homeserver to verify e-mail addresses or to support password resets via e-mail, please configure your homeserver with SMTP access so that it can send e-mails on its own behalf.
 [Consult the configuration documentation for more information.](https://matrix-org.github.io/synapse/latest/usage/configuration/config_documentation.html#email)
@@ -427,7 +435,7 @@ If you require your homeserver to verify e-mail addresses or to support password
 Synapse 1.64.0rc2 (2022-07-29)
 ==============================
 
-This RC reintroduces support for `account_threepid_delegates.email`, which was removed in 1.64.0rc1. It remains deprecated and will be removed altogether in Synapse v1.66.0. ([\#13406](https://github.com/matrix-org/synapse/issues/13406))
+This RC reintroduces support for `account_threepid_delegates.email`, which was removed in 1.64.0rc1. It remains deprecated and will be removed altogether in Synapse 1.66.0. ([\#13406](https://github.com/matrix-org/synapse/issues/13406))
 
 
 Synapse 1.64.0rc1 (2022-07-26)
@@ -676,7 +684,7 @@ Bugfixes
 - Fix a bug introduced in Synapse 1.58 where Synapse would not report full version information when installed from a git checkout. This is a best-effort affair and not guaranteed to be stable. ([\#12973](https://github.com/matrix-org/synapse/issues/12973))
 - Fix a bug introduced in Synapse 1.60 where Synapse would fail to start if the `sqlite3` module was not available. ([\#12979](https://github.com/matrix-org/synapse/issues/12979))
 - Fix a bug where non-standard information was required when requesting the `/hierarchy` API over federation. Introduced
-  in Synapse v1.41.0. ([\#12991](https://github.com/matrix-org/synapse/issues/12991))
+  in Synapse 1.41.0. ([\#12991](https://github.com/matrix-org/synapse/issues/12991))
 - Fix a long-standing bug which meant that rate limiting was not restrictive enough in some cases. ([\#13018](https://github.com/matrix-org/synapse/issues/13018))
 - Fix a bug introduced in Synapse 1.58 where profile requests for a malformed user ID would ccause an internal error. Synapse now returns 400 Bad Request in this situation. ([\#13041](https://github.com/matrix-org/synapse/issues/13041))
 - Fix some inconsistencies in the event authentication code. ([\#13087](https://github.com/matrix-org/synapse/issues/13087), [\#13088](https://github.com/matrix-org/synapse/issues/13088))
@@ -1269,7 +1277,7 @@ If you have already upgraded to Synapse 1.57.0 without problem, then you have no
 Updates to the Docker image
 ---------------------------
 
-- Include version 0.2.0 of the Synapse LDAP Auth Provider module in the Docker image. This matches the version that was present in the Docker image for Synapse v1.56.0. ([\#12512](https://github.com/matrix-org/synapse/issues/12512))
+- Include version 0.2.0 of the Synapse LDAP Auth Provider module in the Docker image. This matches the version that was present in the Docker image for Synapse 1.56.0. ([\#12512](https://github.com/matrix-org/synapse/issues/12512))
 
 
 Synapse 1.57.0 (2022-04-19)
@@ -1521,10 +1529,10 @@ Features
 Bugfixes
 --------
 
-- Use the proper serialization format for bundled thread aggregations. The bug has existed since Synapse v1.48.0. ([\#12090](https://github.com/matrix-org/synapse/issues/12090))
+- Use the proper serialization format for bundled thread aggregations. The bug has existed since Synapse 1.48.0. ([\#12090](https://github.com/matrix-org/synapse/issues/12090))
 - Fix a long-standing bug when redacting events with relations. ([\#12113](https://github.com/matrix-org/synapse/issues/12113), [\#12121](https://github.com/matrix-org/synapse/issues/12121), [\#12130](https://github.com/matrix-org/synapse/issues/12130), [\#12189](https://github.com/matrix-org/synapse/issues/12189))
 - Fix a bug introduced in Synapse 1.7.2 whereby background updates are never run with the default background batch size. ([\#12157](https://github.com/matrix-org/synapse/issues/12157))
-- Fix a bug where non-standard information was returned from the `/hierarchy` API. Introduced in Synapse v1.41.0. ([\#12175](https://github.com/matrix-org/synapse/issues/12175))
+- Fix a bug where non-standard information was returned from the `/hierarchy` API. Introduced in Synapse 1.41.0. ([\#12175](https://github.com/matrix-org/synapse/issues/12175))
 - Fix a bug introduced in Synapse 1.54.0 that broke background updates on sqlite homeservers while search was disabled. ([\#12215](https://github.com/matrix-org/synapse/issues/12215))
 - Fix a long-standing bug when a `filter` argument with `event_fields` which did not include the `unsigned` field could result in a 500 error on `/sync`. ([\#12234](https://github.com/matrix-org/synapse/issues/12234))
 
@@ -1909,15 +1917,15 @@ Bugfixes
 - Fix a long-standing issue which could cause Synapse to incorrectly accept data in the unsigned field of events
   received over federation. ([\#11530](https://github.com/matrix-org/synapse/issues/11530))
 - Fix a long-standing bug where Synapse wouldn't cache a response indicating that a remote user has no devices. ([\#11587](https://github.com/matrix-org/synapse/issues/11587))
-- Fix an error that occurs whilst trying to get the federation status of a destination server that was working normally. This admin API was newly introduced in Synapse v1.49.0. ([\#11593](https://github.com/matrix-org/synapse/issues/11593))
+- Fix an error that occurs whilst trying to get the federation status of a destination server that was working normally. This admin API was newly introduced in Synapse 1.49.0. ([\#11593](https://github.com/matrix-org/synapse/issues/11593))
 - Fix bundled aggregations not being included in the `/sync` response, per [MSC2675](https://github.com/matrix-org/matrix-doc/pull/2675). ([\#11612](https://github.com/matrix-org/synapse/issues/11612), [\#11659](https://github.com/matrix-org/synapse/issues/11659), [\#11791](https://github.com/matrix-org/synapse/issues/11791))
 - Fix the `/_matrix/client/v1/room/{roomId}/hierarchy` endpoint returning incorrect fields which have been present since Synapse 1.49.0. ([\#11667](https://github.com/matrix-org/synapse/issues/11667))
 - Fix preview of some GIF URLs (like tenor.com). Contributed by Philippe Daouadi. ([\#11669](https://github.com/matrix-org/synapse/issues/11669))
-- Fix a bug where only the first 50 rooms from a space were returned from the `/hierarchy` API. This has existed since the introduction of the API in Synapse v1.41.0. ([\#11695](https://github.com/matrix-org/synapse/issues/11695))
-- Fix a bug introduced in Synapse v1.18.0 where password reset and address validation emails would not be sent if their subject was configured to use the 'app' template variable. Contributed by @br4nnigan. ([\#11710](https://github.com/matrix-org/synapse/issues/11710), [\#11745](https://github.com/matrix-org/synapse/issues/11745))
+- Fix a bug where only the first 50 rooms from a space were returned from the `/hierarchy` API. This has existed since the introduction of the API in Synapse 1.41.0. ([\#11695](https://github.com/matrix-org/synapse/issues/11695))
+- Fix a bug introduced in Synapse 1.18.0 where password reset and address validation emails would not be sent if their subject was configured to use the 'app' template variable. Contributed by @br4nnigan. ([\#11710](https://github.com/matrix-org/synapse/issues/11710), [\#11745](https://github.com/matrix-org/synapse/issues/11745))
 - Make the 'List Rooms' Admin API sort stable. Contributed by Daniël Sonck. ([\#11737](https://github.com/matrix-org/synapse/issues/11737))
 - Fix a long-standing bug where space hierarchy over federation would only work correctly some of the time. ([\#11775](https://github.com/matrix-org/synapse/issues/11775))
-- Fix a bug introduced in Synapse v1.46.0 that prevented `on_logged_out` module callbacks from being correctly awaited by Synapse. ([\#11786](https://github.com/matrix-org/synapse/issues/11786))
+- Fix a bug introduced in Synapse 1.46.0 that prevented `on_logged_out` module callbacks from being correctly awaited by Synapse. ([\#11786](https://github.com/matrix-org/synapse/issues/11786))
 
 
 Improved Documentation
@@ -1997,8 +2005,8 @@ This release candidate fixes a federation-breaking regression introduced in Syna
 Bugfixes
 --------
 
-- Fix a bug introduced in Synapse v1.0.0 whereby some device list updates would not be sent to remote homeservers if there were too many to send at once. ([\#11729](https://github.com/matrix-org/synapse/issues/11729))
-- Fix a bug introduced in Synapse v1.50.0rc1 whereby outbound federation could fail because too many EDUs were produced for device updates. ([\#11730](https://github.com/matrix-org/synapse/issues/11730))
+- Fix a bug introduced in Synapse 1.0.0 whereby some device list updates would not be sent to remote homeservers if there were too many to send at once. ([\#11729](https://github.com/matrix-org/synapse/issues/11729))
+- Fix a bug introduced in Synapse 1.50.0rc1 whereby outbound federation could fail because too many EDUs were produced for device updates. ([\#11730](https://github.com/matrix-org/synapse/issues/11730))
 
 
 Improved Documentation
diff --git a/changelog.d/13487.misc b/changelog.d/13487.misc
new file mode 100644
index 0000000000..761adc8b05
--- /dev/null
+++ b/changelog.d/13487.misc
@@ -0,0 +1 @@
+Speed up creation of DM rooms.
diff --git a/changelog.d/13556.feature b/changelog.d/13556.feature
new file mode 100644
index 0000000000..f9d63db6c0
--- /dev/null
+++ b/changelog.d/13556.feature
@@ -0,0 +1 @@
+Allow server admins to require a manual approval process before new accounts can be used (using [MSC3866](https://github.com/matrix-org/matrix-spec-proposals/pull/3866)).
diff --git a/changelog.d/13719.bugfix b/changelog.d/13719.bugfix
new file mode 100644
index 0000000000..4318f4daff
--- /dev/null
+++ b/changelog.d/13719.bugfix
@@ -0,0 +1 @@
+Send invite push notifications for invite over federation.
diff --git a/changelog.d/13787.misc b/changelog.d/13787.misc
new file mode 100644
index 0000000000..a9b93717f0
--- /dev/null
+++ b/changelog.d/13787.misc
@@ -0,0 +1 @@
+Optimise get rooms for user calls. Contributed by Nick @ Beeper (@fizzadar).
diff --git a/changelog.d/13800.misc b/changelog.d/13800.misc
new file mode 100644
index 0000000000..761adc8b05
--- /dev/null
+++ b/changelog.d/13800.misc
@@ -0,0 +1 @@
+Speed up creation of DM rooms.
diff --git a/changelog.d/13838.misc b/changelog.d/13838.misc
new file mode 100644
index 0000000000..28bddb7059
--- /dev/null
+++ b/changelog.d/13838.misc
@@ -0,0 +1 @@
+Port push rules to using Rust.
diff --git a/changelog.d/13839.misc b/changelog.d/13839.misc
new file mode 100644
index 0000000000..549872c90f
--- /dev/null
+++ b/changelog.d/13839.misc
@@ -0,0 +1 @@
+Carry IdP Session IDs through user-mapping sessions.
diff --git a/changelog.d/13863.bugfix b/changelog.d/13863.bugfix
new file mode 100644
index 0000000000..74264a4fab
--- /dev/null
+++ b/changelog.d/13863.bugfix
@@ -0,0 +1 @@
+Fix `have_seen_event` cache not being invalidated after we persist an event which causes inefficiency effects like extra `/state` federation calls.
diff --git a/changelog.d/13867.misc b/changelog.d/13867.misc
new file mode 100644
index 0000000000..1205214598
--- /dev/null
+++ b/changelog.d/13867.misc
@@ -0,0 +1 @@
+Correct the comments in the complement dockerfile.
diff --git a/changelog.d/13868.misc b/changelog.d/13868.misc
new file mode 100644
index 0000000000..d7a99c042a
--- /dev/null
+++ b/changelog.d/13868.misc
@@ -0,0 +1 @@
+Fix unstable MSC3882 endpoint being incorrectly available on stable API versions.
\ No newline at end of file
diff --git a/changelog.d/13872.bugfix b/changelog.d/13872.bugfix
new file mode 100644
index 0000000000..67d3d9e643
--- /dev/null
+++ b/changelog.d/13872.bugfix
@@ -0,0 +1 @@
+Faster room joins: Fix a bug introduced in 1.66.0 where an error would be logged when syncing after joining a room.
diff --git a/changelog.d/13879.misc b/changelog.d/13879.misc
new file mode 100644
index 0000000000..3cc2a2420f
--- /dev/null
+++ b/changelog.d/13879.misc
@@ -0,0 +1 @@
+Only pull relevant backfill points from the database based on the current depth and limit (instead of all) every time we want to `/backfill`.
diff --git a/changelog.d/13885.misc b/changelog.d/13885.misc
new file mode 100644
index 0000000000..bc76b862df
--- /dev/null
+++ b/changelog.d/13885.misc
@@ -0,0 +1 @@
+Correctly handle a race with device lists when a remote user leaves during a partial join.
diff --git a/changelog.d/13890.misc b/changelog.d/13890.misc
new file mode 100644
index 0000000000..bf76cf7be7
--- /dev/null
+++ b/changelog.d/13890.misc
@@ -0,0 +1 @@
+Improve backfill robustness by trying more servers when we get a `4xx` error back.
\ No newline at end of file
diff --git a/changelog.d/13892.feature b/changelog.d/13892.feature
new file mode 100644
index 0000000000..df3f576536
--- /dev/null
+++ b/changelog.d/13892.feature
@@ -0,0 +1 @@
+Faster remote room joins: record _when_ we first partial-join to a room.
diff --git a/changelog.d/13904.bugfix b/changelog.d/13904.bugfix
new file mode 100644
index 0000000000..397a3108ac
--- /dev/null
+++ b/changelog.d/13904.bugfix
@@ -0,0 +1 @@
+Fix a bug introduced in 1.66 where some required fields in the pushrules sent to clients were not present anymore. Contributed by Nico.
diff --git a/changelog.d/13913.misc b/changelog.d/13913.misc
new file mode 100644
index 0000000000..30b4401049
--- /dev/null
+++ b/changelog.d/13913.misc
@@ -0,0 +1 @@
+Faster remote room joins: correctly handle remote device list updates during a partial join.
diff --git a/changelog.d/13914.misc b/changelog.d/13914.misc
new file mode 100644
index 0000000000..c29bc25d38
--- /dev/null
+++ b/changelog.d/13914.misc
@@ -0,0 +1 @@
+Complement image: propagate SIGTERM to all workers.
diff --git a/changelog.d/13915.doc b/changelog.d/13915.doc
new file mode 100644
index 0000000000..828cc30536
--- /dev/null
+++ b/changelog.d/13915.doc
@@ -0,0 +1 @@
+Emphasize the right reasons when to use `(room_id, event_id)` in a database schema.
diff --git a/changelog.d/13920.feature b/changelog.d/13920.feature
new file mode 100644
index 0000000000..aee702bcd2
--- /dev/null
+++ b/changelog.d/13920.feature
@@ -0,0 +1 @@
+Support a `dir` parameter on the `/relations` endpoint per [MSC3715](https://github.com/matrix-org/matrix-doc/pull/3715).
diff --git a/changelog.d/13922.bugfix b/changelog.d/13922.bugfix
new file mode 100644
index 0000000000..7269d28dee
--- /dev/null
+++ b/changelog.d/13922.bugfix
@@ -0,0 +1 @@
+Fix long-standing bug where device updates could cause delays sending out to-device messages over federation.
diff --git a/changelog.d/13924.misc b/changelog.d/13924.misc
new file mode 100644
index 0000000000..7770b6f03f
--- /dev/null
+++ b/changelog.d/13924.misc
@@ -0,0 +1 @@
+Update an innaccurate comment in Synapse's upsert database helper.
diff --git a/changelog.d/13925.misc b/changelog.d/13925.misc
new file mode 100644
index 0000000000..f490ab122e
--- /dev/null
+++ b/changelog.d/13925.misc
@@ -0,0 +1 @@
+Update mypy (0.950 -> 0.981) and mypy-zope (0.3.7 -> 0.3.11).
diff --git a/changelog.d/13928.doc b/changelog.d/13928.doc
new file mode 100644
index 0000000000..04cd06f19d
--- /dev/null
+++ b/changelog.d/13928.doc
@@ -0,0 +1 @@
+Add instruction to contributing guide for running unit tests in parallel. Contributed by @ashfame.
diff --git a/changelog.d/13930.doc b/changelog.d/13930.doc
new file mode 100644
index 0000000000..7cc3206501
--- /dev/null
+++ b/changelog.d/13930.doc
@@ -0,0 +1 @@
+Update the man page for the `hash_password` script to correct the default number of bcrypt rounds performed.
\ No newline at end of file
diff --git a/changelog.d/13931.doc b/changelog.d/13931.doc
new file mode 100644
index 0000000000..85e74fbb3b
--- /dev/null
+++ b/changelog.d/13931.doc
@@ -0,0 +1 @@
+Clarify that the `auto_join_rooms` config option can also be used with Space aliases.
\ No newline at end of file
diff --git a/changelog.d/13932.feature b/changelog.d/13932.feature
new file mode 100644
index 0000000000..d0cb902dff
--- /dev/null
+++ b/changelog.d/13932.feature
@@ -0,0 +1 @@
+Experimental support for thread-specific receipts ([MSC3771](https://github.com/matrix-org/matrix-spec-proposals/pull/3771)).
diff --git a/changelog.d/13934.misc b/changelog.d/13934.misc
new file mode 100644
index 0000000000..6610a9f567
--- /dev/null
+++ b/changelog.d/13934.misc
@@ -0,0 +1 @@
+Correctly handle sending local device list updates to remote servers during a partial join.
diff --git a/changelog.d/13936.feature b/changelog.d/13936.feature
new file mode 100644
index 0000000000..d86bf7ed80
--- /dev/null
+++ b/changelog.d/13936.feature
@@ -0,0 +1 @@
+Exponentially backoff from backfilling the same event over and over.
diff --git a/changelog.d/13937.feature b/changelog.d/13937.feature
new file mode 100644
index 0000000000..d0cb902dff
--- /dev/null
+++ b/changelog.d/13937.feature
@@ -0,0 +1 @@
+Experimental support for thread-specific receipts ([MSC3771](https://github.com/matrix-org/matrix-spec-proposals/pull/3771)).
diff --git a/changelog.d/13939.feature b/changelog.d/13939.feature
new file mode 100644
index 0000000000..d0cb902dff
--- /dev/null
+++ b/changelog.d/13939.feature
@@ -0,0 +1 @@
+Experimental support for thread-specific receipts ([MSC3771](https://github.com/matrix-org/matrix-spec-proposals/pull/3771)).
diff --git a/changelog.d/13947.feature b/changelog.d/13947.feature
new file mode 100644
index 0000000000..a0b3cfe18c
--- /dev/null
+++ b/changelog.d/13947.feature
@@ -0,0 +1 @@
+Add cache invalidation across workers to module API.
diff --git a/changelog.d/13952.bugfix b/changelog.d/13952.bugfix
new file mode 100644
index 0000000000..a6af20f051
--- /dev/null
+++ b/changelog.d/13952.bugfix
@@ -0,0 +1 @@
+Fix a bug introduced in v1.68.0 where Synapse would require `setuptools_rust` at runtime, even though the package is only required at build time.
diff --git a/changelog.d/13957.feature b/changelog.d/13957.feature
new file mode 100644
index 0000000000..4080147357
--- /dev/null
+++ b/changelog.d/13957.feature
@@ -0,0 +1 @@
+Ask mail servers receiving emails from Synapse to not send automatic reply (e.g. out-of-office responses).
diff --git a/changelog.d/13966.misc b/changelog.d/13966.misc
new file mode 100644
index 0000000000..b54ad5c776
--- /dev/null
+++ b/changelog.d/13966.misc
@@ -0,0 +1 @@
+Refactor language in user directory `_track_user_joined_room` code to make it more clear that we use both local and remote users.
diff --git a/changelog.d/13972.bugfix b/changelog.d/13972.bugfix
new file mode 100644
index 0000000000..4c1e19ef8c
--- /dev/null
+++ b/changelog.d/13972.bugfix
@@ -0,0 +1 @@
+Fix a performance regression in the `get_users_in_room` database query. Introduced in v1.67.0.
diff --git a/changelog.d/13973.misc b/changelog.d/13973.misc
new file mode 100644
index 0000000000..58150a2b35
--- /dev/null
+++ b/changelog.d/13973.misc
@@ -0,0 +1 @@
+Speed up calculating push actions in large rooms.
diff --git a/changelog.d/13974.doc b/changelog.d/13974.doc
new file mode 100644
index 0000000000..c4ab17db53
--- /dev/null
+++ b/changelog.d/13974.doc
@@ -0,0 +1 @@
+Add some cross references to worker documentation.
diff --git a/changelog.d/13976.misc b/changelog.d/13976.misc
new file mode 100644
index 0000000000..c235d3a4ac
--- /dev/null
+++ b/changelog.d/13976.misc
@@ -0,0 +1 @@
+Enable update notifications from Github's dependabot.
diff --git a/changelog.d/13992.misc b/changelog.d/13992.misc
new file mode 100644
index 0000000000..58150a2b35
--- /dev/null
+++ b/changelog.d/13992.misc
@@ -0,0 +1 @@
+Speed up calculating push actions in large rooms.
diff --git a/changelog.d/13993.misc b/changelog.d/13993.misc
new file mode 100644
index 0000000000..f490ab122e
--- /dev/null
+++ b/changelog.d/13993.misc
@@ -0,0 +1 @@
+Update mypy (0.950 -> 0.981) and mypy-zope (0.3.7 -> 0.3.11).
diff --git a/debian/changelog b/debian/changelog
index d6c7639775..01fa49aa05 100644
--- a/debian/changelog
+++ b/debian/changelog
@@ -5,6 +5,12 @@ matrix-synapse-py3 (1.69.0~rc1+nmu1) UNRELEASED; urgency=medium
 
  -- Synapse Packaging team <packages@matrix.org>  Mon, 26 Sep 2022 18:05:09 +0100
 
+matrix-synapse-py3 (1.68.0) stable; urgency=medium
+
+  * New Synapse release 1.68.0.
+
+ -- Synapse Packaging team <packages@matrix.org>  Tue, 27 Sep 2022 12:02:09 +0100
+
 matrix-synapse-py3 (1.68.0~rc2) stable; urgency=medium
 
   * New Synapse release 1.68.0rc2.
diff --git a/debian/hash_password.1 b/debian/hash_password.1
index d64b91e7c8..39fa3ffcbf 100644
--- a/debian/hash_password.1
+++ b/debian/hash_password.1
@@ -10,7 +10,7 @@
 .P
 \fBhash_password\fR takes a password as an parameter either on the command line or the \fBSTDIN\fR if not supplied\.
 .P
-It accepts an YAML file which can be used to specify parameters like the number of rounds for bcrypt and password_config section having the pepper value used for the hashing\. By default \fBbcrypt_rounds\fR is set to \fB10\fR\.
+It accepts an YAML file which can be used to specify parameters like the number of rounds for bcrypt and password_config section having the pepper value used for the hashing\. By default \fBbcrypt_rounds\fR is set to \fB12\fR\.
 .P
 The hashed password is written on the \fBSTDOUT\fR\.
 .SH "FILES"
diff --git a/docker/complement/Dockerfile b/docker/complement/Dockerfile
index 0e13722d1c..c0935c99a8 100644
--- a/docker/complement/Dockerfile
+++ b/docker/complement/Dockerfile
@@ -8,19 +8,15 @@
 
 ARG SYNAPSE_VERSION=latest
 
-# first of all, we create a base image with a postgres server and database,
-# which we can copy into the target image. For repeated rebuilds, this is
-# much faster than apt installing postgres each time.
-#
-# This trick only works because (a) the Synapse image happens to have all the
-# shared libraries that postgres wants, (b) we use a postgres image based on
-# the same debian version as Synapse's docker image (so the versions of the
-# shared libraries match).
-
-# now build the final image, based on the Synapse image.
-
 FROM matrixdotorg/synapse-workers:$SYNAPSE_VERSION
-    # copy the postgres installation over from the image we built above
+    # First of all, we copy postgres server from the official postgres image,
+    # since for repeated rebuilds, this is much faster than apt installing
+    # postgres each time.
+
+    # This trick only works because (a) the Synapse image happens to have all the
+    # shared libraries that postgres wants, (b) we use a postgres image based on
+    # the same debian version as Synapse's docker image (so the versions of the
+    # shared libraries match).
     RUN adduser --system --uid 999 postgres --home /var/lib/postgresql
     COPY --from=postgres:13-bullseye /usr/lib/postgresql /usr/lib/postgresql
     COPY --from=postgres:13-bullseye /usr/share/postgresql /usr/share/postgresql
@@ -28,7 +24,7 @@ FROM matrixdotorg/synapse-workers:$SYNAPSE_VERSION
     ENV PATH="${PATH}:/usr/lib/postgresql/13/bin"
     ENV PGDATA=/var/lib/postgresql/data
 
-    # initialise the database cluster in /var/lib/postgresql
+    # We also initialize the database at build time, rather than runtime, so that it's faster to spin up the image.
     RUN gosu postgres initdb --locale=C --encoding=UTF-8 --auth-host password
 
     # Configure a password and create a database for Synapse
diff --git a/docs/development/contributing_guide.md b/docs/development/contributing_guide.md
index cb0d727efa..5c37225168 100644
--- a/docs/development/contributing_guide.md
+++ b/docs/development/contributing_guide.md
@@ -167,6 +167,12 @@ was broken. They are slower than the linters but will typically catch more error
 poetry run trial tests
 ```
 
+You can run unit tests in parallel by specifying `-jX` argument to `trial` where `X` is the number of parallel runners you want. To use 4 cpu cores, you would run them like:
+
+```sh
+poetry run trial -j4 tests
+```
+
 If you wish to only run *some* unit tests, you may specify
 another module instead of `tests` - or a test class or a method:
 
diff --git a/docs/development/database_schema.md b/docs/development/database_schema.md
index e9b925ddd8..29945c264e 100644
--- a/docs/development/database_schema.md
+++ b/docs/development/database_schema.md
@@ -195,23 +195,24 @@ There are three separate aspects to this:
 
 ## `event_id` global uniqueness
 
-In room versions `1` and `2` it's possible to end up with two events with the
-same `event_id` (in the same or different rooms). After room version `3`, that
-can only happen with a hash collision, which we basically hope will never
-happen.
-
-There are several places in Synapse and even Matrix APIs like [`GET
+`event_id`'s can be considered globally unique although there has been a lot of
+debate on this topic in places like
+[MSC2779](https://github.com/matrix-org/matrix-spec-proposals/issues/2779) and
+[MSC2848](https://github.com/matrix-org/matrix-spec-proposals/pull/2848) which
+has no resolution yet (as of 2022-09-01). There are several places in Synapse
+and even in the Matrix APIs like [`GET
 /_matrix/federation/v1/event/{eventId}`](https://spec.matrix.org/v1.1/server-server-api/#get_matrixfederationv1eventeventid)
 where we assume that event IDs are globally unique.
 
-But hash collisions are still possible, and by treating event IDs as room
-scoped, we can reduce the possibility of a hash collision. When scoping
-`event_id` in the database schema, it should be also accompanied by `room_id`
-(`PRIMARY KEY (room_id, event_id)`) and lookups should be done through the pair
-`(room_id, event_id)`.
+When scoping `event_id` in a database schema, it is often nice to accompany it
+with `room_id` (`PRIMARY KEY (room_id, event_id)` and a `FOREIGN KEY(room_id)
+REFERENCES rooms(room_id)`) which makes flexible lookups easy. For example it
+makes it very easy to find and clean up everything in a room when it needs to be
+purged (no need to use sub-`select` query or join from the `events` table).
+
+A note on collisions: In room versions `1` and `2` it's possible to end up with
+two events with the same `event_id` (in the same or different rooms). After room
+version `3`, that can only happen with a hash collision, which we basically hope
+will never happen (SHA256 has a massive big key space).
 
-There has been a lot of debate on this in places like
-https://github.com/matrix-org/matrix-spec-proposals/issues/2779 and
-[MSC2848](https://github.com/matrix-org/matrix-spec-proposals/pull/2848) which
-has no resolution yet (as of 2022-09-01).
 
diff --git a/docs/systemd-with-workers/workers/media_worker.yaml b/docs/systemd-with-workers/workers/media_worker.yaml
new file mode 100644
index 0000000000..eb34d12492
--- /dev/null
+++ b/docs/systemd-with-workers/workers/media_worker.yaml
@@ -0,0 +1,14 @@
+worker_app: synapse.app.media_repository
+worker_name: media_worker
+
+# The replication listener on the main synapse process.
+worker_replication_host: 127.0.0.1
+worker_replication_http_port: 9093
+
+worker_listeners:
+  - type: http
+    port: 8085
+    resources:
+      - names: [media]
+
+worker_log_config: /etc/matrix-synapse/media-worker-log.yaml
diff --git a/docs/upgrade.md b/docs/upgrade.md
index 7d4c2392e1..c4db19e23d 100644
--- a/docs/upgrade.md
+++ b/docs/upgrade.md
@@ -15,9 +15,8 @@ this document.
     The website <https://endoflife.date> also offers convenient
     summaries.
 
--   If Synapse was installed using [prebuilt
-    packages](setup/installation.md#prebuilt-packages), you will need to follow the
-    normal process for upgrading those packages.
+-   If Synapse was installed using [prebuilt packages](setup/installation.md#prebuilt-packages),
+    you will need to follow the normal process for upgrading those packages.
 
 -   If Synapse was installed using pip then upgrade to the latest
     version by running:
@@ -89,12 +88,48 @@ process, for example:
     dpkg -i matrix-synapse-py3_1.3.0+stretch1_amd64.deb
     ```
 
+# Upgrading to v1.69.0
+
+## Changes to the receipts replication streams
+
+Synapse now includes information indicating if a receipt applies to a thread when
+replicating it to other workers. This is a forwards- and backwards-incompatible
+change: v1.68 and workers cannot process receipts replicated by v1.69 workers, and
+vice versa.
+
+Once all workers are upgraded to v1.69 (or downgraded to v1.68), receipts
+replication will resume as normal.
+
 # Upgrading to v1.68.0
 
-As announced in the upgrade notes for v1.67.0, Synapse now requires a SQLite
-version of 3.27.0 or higher if SQLite is in use and source checkouts of Synapse
-now require a recent Rust compiler.
+Two changes announced in the upgrade notes for v1.67.0 have now landed in v1.68.0.
+
+## SQLite version requirement
+
+Synapse now requires a SQLite version of 3.27.0 or higher if SQLite is configured as
+Synapse's database.
+
+Installations using
+
+- Docker images [from `matrixdotorg`](https://hub.docker.com/r/matrixdotorg/synapse),
+- Debian packages [from Matrix.org](https://packages.matrix.org/), or
+- a PostgreSQL database
+
+are not affected.
+
+## Rust requirement when building from source.
+
+Building from a source checkout of Synapse now requires a recent Rust compiler
+(currently Rust 1.58.1, but see also the
+[Platform Dependency Policy](https://matrix-org.github.io/synapse/latest/deprecation_policy.html)).
+
+Installations using
+
+- Docker images [from `matrixdotorg`](https://hub.docker.com/r/matrixdotorg/synapse),
+- Debian packages [from Matrix.org](https://packages.matrix.org/), or
+- PyPI wheels via `pip install matrix-synapse` (on supported platforms and architectures)
 
+will not be affected.
 
 # Upgrading to v1.67.0
 
@@ -128,12 +163,12 @@ The simplest way of installing Rust is via [rustup.rs](https://rustup.rs/)
 
 ## SQLite version requirement in the next release
 
-From the next major release (v1.68.0) Synapse will require SQLite 3.27.0 or 
+From the next major release (v1.68.0) Synapse will require SQLite 3.27.0 or
 higher. Synapse v1.67.0 will be the last major release supporting SQLite
 versions 3.22 to 3.26.
 
 Those using Docker images or Debian packages from Matrix.org will not be
-affected. If you have installed from source, you should check the version of 
+affected. If you have installed from source, you should check the version of
 SQLite used by Python with:
 
 ```shell
diff --git a/docs/usage/configuration/config_documentation.md b/docs/usage/configuration/config_documentation.md
index 77069f28a7..3510bb45d9 100644
--- a/docs/usage/configuration/config_documentation.md
+++ b/docs/usage/configuration/config_documentation.md
@@ -2230,6 +2230,9 @@ homeserver. If the room already exists, make certain it is a publicly joinable
 room, i.e. the join rule of the room must be set to 'public'. You can find more options
 relating to auto-joining rooms below.
 
+As Spaces are just rooms under the hood, Space aliases may also be
+used.
+
 Example configuration:
 ```yaml
 auto_join_rooms:
@@ -2241,7 +2244,7 @@ auto_join_rooms:
 
 Where `auto_join_rooms` are specified, setting this flag ensures that
 the rooms exist by creating them when the first user on the
-homeserver registers.
+homeserver registers. This option will not create Spaces.
 
 By default the auto-created rooms are publicly joinable from any federated
 server. Use the `autocreate_auto_join_rooms_federated` and
@@ -2259,7 +2262,7 @@ autocreate_auto_join_rooms: false
 ---
 ### `autocreate_auto_join_rooms_federated`
 
-Whether the rooms listen in `auto_join_rooms` that are auto-created are available
+Whether the rooms listed in `auto_join_rooms` that are auto-created are available
 via federation. Only has an effect if `autocreate_auto_join_rooms` is true.
 
 Note that whether a room is federated cannot be modified after
diff --git a/docs/workers.md b/docs/workers.md
index 40b1852313..25f2e13237 100644
--- a/docs/workers.md
+++ b/docs/workers.md
@@ -93,7 +93,6 @@ listener" for the main process; and secondly, you need to enable redis-based
 replication. Optionally, a shared secret can be used to authenticate HTTP
 traffic between workers. For example:
 
-
 ```yaml
 # extend the existing `listeners` section. This defines the ports that the
 # main process will listen on.
@@ -129,7 +128,8 @@ In the config file for each worker, you must specify:
  * The HTTP replication endpoint that it should talk to on the main synapse process
    (`worker_replication_host` and `worker_replication_http_port`)
  * If handling HTTP requests, a `worker_listeners` option with an `http`
-   listener, in the same way as the `listeners` option in the shared config.
+   listener, in the same way as the [`listeners`](usage/configuration/config_documentation.md#listeners)
+   option in the shared config.
  * If handling the `^/_matrix/client/v3/keys/upload` endpoint, the HTTP URI for
    the main process (`worker_main_http_uri`).
 
@@ -285,8 +285,9 @@ For multiple workers not handling the SSO endpoints properly, see
 [#7530](https://github.com/matrix-org/synapse/issues/7530) and
 [#9427](https://github.com/matrix-org/synapse/issues/9427).
 
-Note that a HTTP listener with `client` and `federation` resources must be
-configured in the `worker_listeners` option in the worker config.
+Note that a [HTTP listener](usage/configuration/config_documentation.md#listeners)
+with `client` and `federation` `resources` must be configured in the `worker_listeners`
+option in the worker config.
 
 #### Load balancing
 
@@ -326,7 +327,8 @@ effects of bursts of events from that bridge on events sent by normal users.
 Additionally, the writing of specific streams (such as events) can be moved off
 of the main process to a particular worker.
 
-To enable this, the worker must have a HTTP replication listener configured,
+To enable this, the worker must have a
+[HTTP `replication` listener](usage/configuration/config_documentation.md#listeners) configured,
 have a `worker_name` and be listed in the `instance_map` config. The same worker
 can handle multiple streams, but unless otherwise documented, each stream can only
 have a single writer.
@@ -410,7 +412,7 @@ the stream writer for the `presence` stream:
 There is also support for moving background tasks to a separate
 worker. Background tasks are run periodically or started via replication. Exactly
 which tasks are configured to run depends on your Synapse configuration (e.g. if
-stats is enabled).
+stats is enabled). This worker doesn't handle any REST endpoints itself.
 
 To enable this, the worker must have a `worker_name` and can be configured to run
 background tasks. For example, to move background tasks to a dedicated worker,
@@ -457,8 +459,8 @@ worker application type.
 #### Notifying Application Services
 
 You can designate one generic worker to send output traffic to Application Services.
-
-Specify its name in the shared configuration as follows:
+Doesn't handle any REST endpoints itself, but you should specify its name in the
+shared configuration as follows:
 
 ```yaml
 notify_appservices_from_worker: worker_name
@@ -536,16 +538,12 @@ file to stop the main synapse running background jobs related to managing the
 media repository. Note that doing so will prevent the main process from being
 able to handle the above endpoints.
 
-In the `media_repository` worker configuration file, configure the http listener to
+In the `media_repository` worker configuration file, configure the
+[HTTP listener](usage/configuration/config_documentation.md#listeners) to
 expose the `media` resource. For example:
 
 ```yaml
-worker_listeners:
- - type: http
-   port: 8085
-   resources:
-     - names:
-       - media
+{{#include systemd-with-workers/workers/media_worker.yaml}}
 ```
 
 Note that if running multiple media repositories they must be on the same server
diff --git a/poetry.lock b/poetry.lock
index a106a972f1..2430c3d4b6 100644
--- a/poetry.lock
+++ b/poetry.lock
@@ -584,11 +584,11 @@ python-versions = "*"
 
 [[package]]
 name = "mypy"
-version = "0.950"
+version = "0.981"
 description = "Optional static typing for Python"
 category = "dev"
 optional = false
-python-versions = ">=3.6"
+python-versions = ">=3.7"
 
 [package.dependencies]
 mypy-extensions = ">=0.4.3"
@@ -611,14 +611,14 @@ python-versions = "*"
 
 [[package]]
 name = "mypy-zope"
-version = "0.3.7"
+version = "0.3.11"
 description = "Plugin for mypy to support zope interfaces"
 category = "dev"
 optional = false
 python-versions = "*"
 
 [package.dependencies]
-mypy = "0.950"
+mypy = "0.981"
 "zope.interface" = "*"
 "zope.schema" = "*"
 
@@ -2266,37 +2266,38 @@ msgpack = [
     {file = "msgpack-1.0.3.tar.gz", hash = "sha256:51fdc7fb93615286428ee7758cecc2f374d5ff363bdd884c7ea622a7a327a81e"},
 ]
 mypy = [
-    {file = "mypy-0.950-cp310-cp310-macosx_10_9_universal2.whl", hash = "sha256:cf9c261958a769a3bd38c3e133801ebcd284ffb734ea12d01457cb09eacf7d7b"},
-    {file = "mypy-0.950-cp310-cp310-macosx_10_9_x86_64.whl", hash = "sha256:b5b5bd0ffb11b4aba2bb6d31b8643902c48f990cc92fda4e21afac658044f0c0"},
-    {file = "mypy-0.950-cp310-cp310-macosx_11_0_arm64.whl", hash = "sha256:5e7647df0f8fc947388e6251d728189cfadb3b1e558407f93254e35abc026e22"},
-    {file = "mypy-0.950-cp310-cp310-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_12_x86_64.manylinux2010_x86_64.whl", hash = "sha256:eaff8156016487c1af5ffa5304c3e3fd183edcb412f3e9c72db349faf3f6e0eb"},
-    {file = "mypy-0.950-cp310-cp310-win_amd64.whl", hash = "sha256:563514c7dc504698fb66bb1cf897657a173a496406f1866afae73ab5b3cdb334"},
-    {file = "mypy-0.950-cp36-cp36m-macosx_10_9_x86_64.whl", hash = "sha256:dd4d670eee9610bf61c25c940e9ade2d0ed05eb44227275cce88701fee014b1f"},
-    {file = "mypy-0.950-cp36-cp36m-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_12_x86_64.manylinux2010_x86_64.whl", hash = "sha256:ca75ecf2783395ca3016a5e455cb322ba26b6d33b4b413fcdedfc632e67941dc"},
-    {file = "mypy-0.950-cp36-cp36m-win_amd64.whl", hash = "sha256:6003de687c13196e8a1243a5e4bcce617d79b88f83ee6625437e335d89dfebe2"},
-    {file = "mypy-0.950-cp37-cp37m-macosx_10_9_x86_64.whl", hash = "sha256:4c653e4846f287051599ed8f4b3c044b80e540e88feec76b11044ddc5612ffed"},
-    {file = "mypy-0.950-cp37-cp37m-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_12_x86_64.manylinux2010_x86_64.whl", hash = "sha256:e19736af56947addedce4674c0971e5dceef1b5ec7d667fe86bcd2b07f8f9075"},
-    {file = "mypy-0.950-cp37-cp37m-win_amd64.whl", hash = "sha256:ef7beb2a3582eb7a9f37beaf38a28acfd801988cde688760aea9e6cc4832b10b"},
-    {file = "mypy-0.950-cp38-cp38-macosx_10_9_universal2.whl", hash = "sha256:0112752a6ff07230f9ec2f71b0d3d4e088a910fdce454fdb6553e83ed0eced7d"},
-    {file = "mypy-0.950-cp38-cp38-macosx_10_9_x86_64.whl", hash = "sha256:ee0a36edd332ed2c5208565ae6e3a7afc0eabb53f5327e281f2ef03a6bc7687a"},
-    {file = "mypy-0.950-cp38-cp38-macosx_11_0_arm64.whl", hash = "sha256:77423570c04aca807508a492037abbd72b12a1fb25a385847d191cd50b2c9605"},
-    {file = "mypy-0.950-cp38-cp38-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_12_x86_64.manylinux2010_x86_64.whl", hash = "sha256:5ce6a09042b6da16d773d2110e44f169683d8cc8687e79ec6d1181a72cb028d2"},
-    {file = "mypy-0.950-cp38-cp38-win_amd64.whl", hash = "sha256:5b231afd6a6e951381b9ef09a1223b1feabe13625388db48a8690f8daa9b71ff"},
-    {file = "mypy-0.950-cp39-cp39-macosx_10_9_universal2.whl", hash = "sha256:0384d9f3af49837baa92f559d3fa673e6d2652a16550a9ee07fc08c736f5e6f8"},
-    {file = "mypy-0.950-cp39-cp39-macosx_10_9_x86_64.whl", hash = "sha256:1fdeb0a0f64f2a874a4c1f5271f06e40e1e9779bf55f9567f149466fc7a55038"},
-    {file = "mypy-0.950-cp39-cp39-macosx_11_0_arm64.whl", hash = "sha256:61504b9a5ae166ba5ecfed9e93357fd51aa693d3d434b582a925338a2ff57fd2"},
-    {file = "mypy-0.950-cp39-cp39-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_12_x86_64.manylinux2010_x86_64.whl", hash = "sha256:a952b8bc0ae278fc6316e6384f67bb9a396eb30aced6ad034d3a76120ebcc519"},
-    {file = "mypy-0.950-cp39-cp39-win_amd64.whl", hash = "sha256:eaea21d150fb26d7b4856766e7addcf929119dd19fc832b22e71d942835201ef"},
-    {file = "mypy-0.950-py3-none-any.whl", hash = "sha256:a4d9898f46446bfb6405383b57b96737dcfd0a7f25b748e78ef3e8c576bba3cb"},
-    {file = "mypy-0.950.tar.gz", hash = "sha256:1b333cfbca1762ff15808a0ef4f71b5d3eed8528b23ea1c3fb50543c867d68de"},
+    {file = "mypy-0.981-cp310-cp310-macosx_10_9_universal2.whl", hash = "sha256:4bc460e43b7785f78862dab78674e62ec3cd523485baecfdf81a555ed29ecfa0"},
+    {file = "mypy-0.981-cp310-cp310-macosx_10_9_x86_64.whl", hash = "sha256:756fad8b263b3ba39e4e204ee53042671b660c36c9017412b43af210ddee7b08"},
+    {file = "mypy-0.981-cp310-cp310-macosx_11_0_arm64.whl", hash = "sha256:a16a0145d6d7d00fbede2da3a3096dcc9ecea091adfa8da48fa6a7b75d35562d"},
+    {file = "mypy-0.981-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:ce65f70b14a21fdac84c294cde75e6dbdabbcff22975335e20827b3b94bdbf49"},
+    {file = "mypy-0.981-cp310-cp310-musllinux_1_1_x86_64.whl", hash = "sha256:6e35d764784b42c3e256848fb8ed1d4292c9fc0098413adb28d84974c095b279"},
+    {file = "mypy-0.981-cp310-cp310-win_amd64.whl", hash = "sha256:e53773073c864d5f5cec7f3fc72fbbcef65410cde8cc18d4f7242dea60dac52e"},
+    {file = "mypy-0.981-cp37-cp37m-macosx_10_9_x86_64.whl", hash = "sha256:6ee196b1d10b8b215e835f438e06965d7a480f6fe016eddbc285f13955cca659"},
+    {file = "mypy-0.981-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:8ad21d4c9d3673726cf986ea1d0c9fb66905258709550ddf7944c8f885f208be"},
+    {file = "mypy-0.981-cp37-cp37m-musllinux_1_1_x86_64.whl", hash = "sha256:d1debb09043e1f5ee845fa1e96d180e89115b30e47c5d3ce53bc967bab53f62d"},
+    {file = "mypy-0.981-cp37-cp37m-win_amd64.whl", hash = "sha256:9f362470a3480165c4c6151786b5379351b790d56952005be18bdbdd4c7ce0ae"},
+    {file = "mypy-0.981-cp38-cp38-macosx_10_9_universal2.whl", hash = "sha256:c9e0efb95ed6ca1654951bd5ec2f3fa91b295d78bf6527e026529d4aaa1e0c30"},
+    {file = "mypy-0.981-cp38-cp38-macosx_10_9_x86_64.whl", hash = "sha256:e178eaffc3c5cd211a87965c8c0df6da91ed7d258b5fc72b8e047c3771317ddb"},
+    {file = "mypy-0.981-cp38-cp38-macosx_11_0_arm64.whl", hash = "sha256:06e1eac8d99bd404ed8dd34ca29673c4346e76dd8e612ea507763dccd7e13c7a"},
+    {file = "mypy-0.981-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:fa38f82f53e1e7beb45557ff167c177802ba7b387ad017eab1663d567017c8ee"},
+    {file = "mypy-0.981-cp38-cp38-musllinux_1_1_x86_64.whl", hash = "sha256:64e1f6af81c003f85f0dfed52db632817dabb51b65c0318ffbf5ff51995bbb08"},
+    {file = "mypy-0.981-cp38-cp38-win_amd64.whl", hash = "sha256:e1acf62a8c4f7c092462c738aa2c2489e275ed386320c10b2e9bff31f6f7e8d6"},
+    {file = "mypy-0.981-cp39-cp39-macosx_10_9_universal2.whl", hash = "sha256:b6ede64e52257931315826fdbfc6ea878d89a965580d1a65638ef77cb551f56d"},
+    {file = "mypy-0.981-cp39-cp39-macosx_10_9_x86_64.whl", hash = "sha256:eb3978b191b9fa0488524bb4ffedf2c573340e8c2b4206fc191d44c7093abfb7"},
+    {file = "mypy-0.981-cp39-cp39-macosx_11_0_arm64.whl", hash = "sha256:77f8fcf7b4b3cc0c74fb33ae54a4cd00bb854d65645c48beccf65fa10b17882c"},
+    {file = "mypy-0.981-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:f64d2ce043a209a297df322eb4054dfbaa9de9e8738291706eaafda81ab2b362"},
+    {file = "mypy-0.981-cp39-cp39-musllinux_1_1_x86_64.whl", hash = "sha256:2ee3dbc53d4df7e6e3b1c68ac6a971d3a4fb2852bf10a05fda228721dd44fae1"},
+    {file = "mypy-0.981-cp39-cp39-win_amd64.whl", hash = "sha256:8e8e49aa9cc23aa4c926dc200ce32959d3501c4905147a66ce032f05cb5ecb92"},
+    {file = "mypy-0.981-py3-none-any.whl", hash = "sha256:794f385653e2b749387a42afb1e14c2135e18daeb027e0d97162e4b7031210f8"},
+    {file = "mypy-0.981.tar.gz", hash = "sha256:ad77c13037d3402fbeffda07d51e3f228ba078d1c7096a73759c9419ea031bf4"},
 ]
 mypy-extensions = [
     {file = "mypy_extensions-0.4.3-py2.py3-none-any.whl", hash = "sha256:090fedd75945a69ae91ce1303b5824f428daf5a028d2f6ab8a299250a846f15d"},
     {file = "mypy_extensions-0.4.3.tar.gz", hash = "sha256:2d82818f5bb3e369420cb3c4060a7970edba416647068eb4c5343488a6c604a8"},
 ]
 mypy-zope = [
-    {file = "mypy-zope-0.3.7.tar.gz", hash = "sha256:9da171e78e8ef7ac8922c86af1a62f1b7f3244f121020bd94a2246bc3f33c605"},
-    {file = "mypy_zope-0.3.7-py3-none-any.whl", hash = "sha256:9c7637d066e4d1bafa0651abc091c752009769098043b236446e6725be2bc9c2"},
+    {file = "mypy-zope-0.3.11.tar.gz", hash = "sha256:d4255f9f04d48c79083bbd4e2fea06513a6ac7b8de06f8c4ce563fd85142ca05"},
+    {file = "mypy_zope-0.3.11-py3-none-any.whl", hash = "sha256:ec080a6508d1f7805c8d2054f9fdd13c849742ce96803519e1fdfa3d3cab7140"},
 ]
 netaddr = [
     {file = "netaddr-0.8.0-py2.py3-none-any.whl", hash = "sha256:9666d0232c32d2656e5e5f8d735f58fd6c7457ce52fc21c98d45f2af78f990ac"},
diff --git a/pyproject.toml b/pyproject.toml
index e669867aa0..444d4ee82d 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -57,7 +57,7 @@ manifest-path = "rust/Cargo.toml"
 
 [tool.poetry]
 name = "matrix-synapse"
-version = "1.68.0rc2"
+version = "1.68.0"
 description = "Homeserver for the Matrix decentralised comms protocol"
 authors = ["Matrix.org Team and Contributors <packages@matrix.org>"]
 license = "Apache-2.0"
diff --git a/rust/Cargo.toml b/rust/Cargo.toml
index 44263bf77e..cffaa5b51b 100644
--- a/rust/Cargo.toml
+++ b/rust/Cargo.toml
@@ -11,7 +11,9 @@ rust-version = "1.58.1"
 
 [lib]
 name = "synapse"
-crate-type = ["cdylib"]
+# We generate a `cdylib` for Python and a standard `lib` for running
+# tests/benchmarks.
+crate-type = ["lib", "cdylib"]
 
 [package.metadata.maturin]
 # This is where we tell maturin where to place the built library.
diff --git a/rust/benches/evaluator.rs b/rust/benches/evaluator.rs
new file mode 100644
index 0000000000..ed411461d1
--- /dev/null
+++ b/rust/benches/evaluator.rs
@@ -0,0 +1,149 @@
+// Copyright 2022 The Matrix.org Foundation C.I.C.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#![feature(test)]
+use synapse::push::{
+    evaluator::PushRuleEvaluator, Condition, EventMatchCondition, FilteredPushRules, PushRules,
+};
+use test::Bencher;
+
+extern crate test;
+
+#[bench]
+fn bench_match_exact(b: &mut Bencher) {
+    let flattened_keys = [
+        ("type".to_string(), "m.text".to_string()),
+        ("room_id".to_string(), "!room:server".to_string()),
+        ("content.body".to_string(), "test message".to_string()),
+    ]
+    .into_iter()
+    .collect();
+
+    let eval = PushRuleEvaluator::py_new(
+        flattened_keys,
+        10,
+        0,
+        Default::default(),
+        Default::default(),
+        true,
+    )
+    .unwrap();
+
+    let condition = Condition::Known(synapse::push::KnownCondition::EventMatch(
+        EventMatchCondition {
+            key: "room_id".into(),
+            pattern: Some("!room:server".into()),
+            pattern_type: None,
+        },
+    ));
+
+    let matched = eval.match_condition(&condition, None, None).unwrap();
+    assert!(matched, "Didn't match");
+
+    b.iter(|| eval.match_condition(&condition, None, None).unwrap());
+}
+
+#[bench]
+fn bench_match_word(b: &mut Bencher) {
+    let flattened_keys = [
+        ("type".to_string(), "m.text".to_string()),
+        ("room_id".to_string(), "!room:server".to_string()),
+        ("content.body".to_string(), "test message".to_string()),
+    ]
+    .into_iter()
+    .collect();
+
+    let eval = PushRuleEvaluator::py_new(
+        flattened_keys,
+        10,
+        0,
+        Default::default(),
+        Default::default(),
+        true,
+    )
+    .unwrap();
+
+    let condition = Condition::Known(synapse::push::KnownCondition::EventMatch(
+        EventMatchCondition {
+            key: "content.body".into(),
+            pattern: Some("test".into()),
+            pattern_type: None,
+        },
+    ));
+
+    let matched = eval.match_condition(&condition, None, None).unwrap();
+    assert!(matched, "Didn't match");
+
+    b.iter(|| eval.match_condition(&condition, None, None).unwrap());
+}
+
+#[bench]
+fn bench_match_word_miss(b: &mut Bencher) {
+    let flattened_keys = [
+        ("type".to_string(), "m.text".to_string()),
+        ("room_id".to_string(), "!room:server".to_string()),
+        ("content.body".to_string(), "test message".to_string()),
+    ]
+    .into_iter()
+    .collect();
+
+    let eval = PushRuleEvaluator::py_new(
+        flattened_keys,
+        10,
+        0,
+        Default::default(),
+        Default::default(),
+        true,
+    )
+    .unwrap();
+
+    let condition = Condition::Known(synapse::push::KnownCondition::EventMatch(
+        EventMatchCondition {
+            key: "content.body".into(),
+            pattern: Some("foobar".into()),
+            pattern_type: None,
+        },
+    ));
+
+    let matched = eval.match_condition(&condition, None, None).unwrap();
+    assert!(!matched, "Didn't match");
+
+    b.iter(|| eval.match_condition(&condition, None, None).unwrap());
+}
+
+#[bench]
+fn bench_eval_message(b: &mut Bencher) {
+    let flattened_keys = [
+        ("type".to_string(), "m.text".to_string()),
+        ("room_id".to_string(), "!room:server".to_string()),
+        ("content.body".to_string(), "test message".to_string()),
+    ]
+    .into_iter()
+    .collect();
+
+    let eval = PushRuleEvaluator::py_new(
+        flattened_keys,
+        10,
+        0,
+        Default::default(),
+        Default::default(),
+        true,
+    )
+    .unwrap();
+
+    let rules =
+        FilteredPushRules::py_new(PushRules::new(Vec::new()), Default::default(), false, false);
+
+    b.iter(|| eval.run(&rules, Some("bob"), Some("person")));
+}
diff --git a/rust/benches/glob.rs b/rust/benches/glob.rs
new file mode 100644
index 0000000000..b6697d9285
--- /dev/null
+++ b/rust/benches/glob.rs
@@ -0,0 +1,40 @@
+// Copyright 2022 The Matrix.org Foundation C.I.C.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#![feature(test)]
+
+use synapse::push::utils::{glob_to_regex, GlobMatchType};
+use test::Bencher;
+
+extern crate test;
+
+#[bench]
+fn bench_whole(b: &mut Bencher) {
+    b.iter(|| glob_to_regex("test", GlobMatchType::Whole));
+}
+
+#[bench]
+fn bench_word(b: &mut Bencher) {
+    b.iter(|| glob_to_regex("test", GlobMatchType::Word));
+}
+
+#[bench]
+fn bench_whole_wildcard_run(b: &mut Bencher) {
+    b.iter(|| glob_to_regex("test***??*?*?foo", GlobMatchType::Whole));
+}
+
+#[bench]
+fn bench_word_wildcard_run(b: &mut Bencher) {
+    b.iter(|| glob_to_regex("test***??*?*?foo", GlobMatchType::Whole));
+}
diff --git a/rust/build.rs b/rust/build.rs
index 2117975e56..ef370e6b41 100644
--- a/rust/build.rs
+++ b/rust/build.rs
@@ -22,7 +22,7 @@ fn main() -> Result<(), std::io::Error> {
 
         for entry in entries {
             if entry.is_dir() {
-                dirs.push(entry)
+                dirs.push(entry);
             } else {
                 paths.push(entry.to_str().expect("valid rust paths").to_string());
             }
diff --git a/rust/src/push/base_rules.rs b/rust/src/push/base_rules.rs
index 7c62bc4849..bb59676bde 100644
--- a/rust/src/push/base_rules.rs
+++ b/rust/src/push/base_rules.rs
@@ -262,6 +262,7 @@ pub const BASE_APPEND_UNDERRIDE_RULES: &[PushRule] = &[
         priority_class: 1,
         conditions: Cow::Borrowed(&[Condition::Known(KnownCondition::RelationMatch {
             rel_type: Cow::Borrowed("m.thread"),
+            event_type_pattern: None,
             sender: None,
             sender_type: Some(Cow::Borrowed("user_id")),
         })]),
diff --git a/rust/src/push/evaluator.rs b/rust/src/push/evaluator.rs
new file mode 100644
index 0000000000..efe88ec76e
--- /dev/null
+++ b/rust/src/push/evaluator.rs
@@ -0,0 +1,374 @@
+// Copyright 2022 The Matrix.org Foundation C.I.C.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+use std::{
+    borrow::Cow,
+    collections::{BTreeMap, BTreeSet},
+};
+
+use anyhow::{Context, Error};
+use lazy_static::lazy_static;
+use log::warn;
+use pyo3::prelude::*;
+use regex::Regex;
+
+use super::{
+    utils::{get_glob_matcher, get_localpart_from_id, GlobMatchType},
+    Action, Condition, EventMatchCondition, FilteredPushRules, KnownCondition,
+};
+
+lazy_static! {
+    /// Used to parse the `is` clause in the room member count condition.
+    static ref INEQUALITY_EXPR: Regex = Regex::new(r"^([=<>]*)([0-9]+)$").expect("valid regex");
+}
+
+/// Allows running a set of push rules against a particular event.
+#[pyclass]
+pub struct PushRuleEvaluator {
+    /// A mapping of "flattened" keys to string values in the event, e.g.
+    /// includes things like "type" and "content.msgtype".
+    flattened_keys: BTreeMap<String, String>,
+
+    /// The "content.body", if any.
+    body: String,
+
+    /// The number of users in the room.
+    room_member_count: u64,
+
+    /// The `notifications` section of the current power levels in the room.
+    notification_power_levels: BTreeMap<String, i64>,
+
+    /// The relations related to the event as a mapping from relation type to
+    /// set of sender/event type 2-tuples.
+    relations: BTreeMap<String, BTreeSet<(String, String)>>,
+
+    /// Is running "relation" conditions enabled?
+    relation_match_enabled: bool,
+
+    /// The power level of the sender of the event, or None if event is an
+    /// outlier.
+    sender_power_level: Option<i64>,
+}
+
+#[pymethods]
+impl PushRuleEvaluator {
+    /// Create a new `PushRuleEvaluator`. See struct docstring for details.
+    #[new]
+    pub fn py_new(
+        flattened_keys: BTreeMap<String, String>,
+        room_member_count: u64,
+        sender_power_level: Option<i64>,
+        notification_power_levels: BTreeMap<String, i64>,
+        relations: BTreeMap<String, BTreeSet<(String, String)>>,
+        relation_match_enabled: bool,
+    ) -> Result<Self, Error> {
+        let body = flattened_keys
+            .get("content.body")
+            .cloned()
+            .unwrap_or_default();
+
+        Ok(PushRuleEvaluator {
+            flattened_keys,
+            body,
+            room_member_count,
+            notification_power_levels,
+            relations,
+            relation_match_enabled,
+            sender_power_level,
+        })
+    }
+
+    /// Run the evaluator with the given push rules, for the given user ID and
+    /// display name of the user.
+    ///
+    /// Passing in None will skip evaluating rules matching user ID and display
+    /// name.
+    ///
+    /// Returns the set of actions, if any, that match (filtering out any
+    /// `dont_notify` actions).
+    pub fn run(
+        &self,
+        push_rules: &FilteredPushRules,
+        user_id: Option<&str>,
+        display_name: Option<&str>,
+    ) -> Vec<Action> {
+        'outer: for (push_rule, enabled) in push_rules.iter() {
+            if !enabled {
+                continue;
+            }
+
+            for condition in push_rule.conditions.iter() {
+                match self.match_condition(condition, user_id, display_name) {
+                    Ok(true) => {}
+                    Ok(false) => continue 'outer,
+                    Err(err) => {
+                        warn!("Condition match failed {err}");
+                        continue 'outer;
+                    }
+                }
+            }
+
+            let actions = push_rule
+                .actions
+                .iter()
+                // Filter out "dont_notify" actions, as we don't store them.
+                .filter(|a| **a != Action::DontNotify)
+                .cloned()
+                .collect();
+
+            return actions;
+        }
+
+        Vec::new()
+    }
+
+    /// Check if the given condition matches.
+    fn matches(
+        &self,
+        condition: Condition,
+        user_id: Option<&str>,
+        display_name: Option<&str>,
+    ) -> bool {
+        match self.match_condition(&condition, user_id, display_name) {
+            Ok(true) => true,
+            Ok(false) => false,
+            Err(err) => {
+                warn!("Condition match failed {err}");
+                false
+            }
+        }
+    }
+}
+
+impl PushRuleEvaluator {
+    /// Match a given `Condition` for a push rule.
+    pub fn match_condition(
+        &self,
+        condition: &Condition,
+        user_id: Option<&str>,
+        display_name: Option<&str>,
+    ) -> Result<bool, Error> {
+        let known_condition = match condition {
+            Condition::Known(known) => known,
+            Condition::Unknown(_) => {
+                return Ok(false);
+            }
+        };
+
+        let result = match known_condition {
+            KnownCondition::EventMatch(event_match) => {
+                self.match_event_match(event_match, user_id)?
+            }
+            KnownCondition::ContainsDisplayName => {
+                if let Some(dn) = display_name {
+                    if !dn.is_empty() {
+                        get_glob_matcher(dn, GlobMatchType::Word)?.is_match(&self.body)?
+                    } else {
+                        // We specifically ignore empty display names, as otherwise
+                        // they would always match.
+                        false
+                    }
+                } else {
+                    false
+                }
+            }
+            KnownCondition::RoomMemberCount { is } => {
+                if let Some(is) = is {
+                    self.match_member_count(is)?
+                } else {
+                    false
+                }
+            }
+            KnownCondition::SenderNotificationPermission { key } => {
+                if let Some(sender_power_level) = &self.sender_power_level {
+                    let required_level = self
+                        .notification_power_levels
+                        .get(key.as_ref())
+                        .copied()
+                        .unwrap_or(50);
+
+                    *sender_power_level >= required_level
+                } else {
+                    false
+                }
+            }
+            KnownCondition::RelationMatch {
+                rel_type,
+                event_type_pattern,
+                sender,
+                sender_type,
+            } => {
+                self.match_relations(rel_type, sender, sender_type, user_id, event_type_pattern)?
+            }
+        };
+
+        Ok(result)
+    }
+
+    /// Evaluates a relation condition.
+    fn match_relations(
+        &self,
+        rel_type: &str,
+        sender: &Option<Cow<str>>,
+        sender_type: &Option<Cow<str>>,
+        user_id: Option<&str>,
+        event_type_pattern: &Option<Cow<str>>,
+    ) -> Result<bool, Error> {
+        // First check if relation matching is enabled...
+        if !self.relation_match_enabled {
+            return Ok(false);
+        }
+
+        // ... and if there are any relations to match against.
+        let relations = if let Some(relations) = self.relations.get(rel_type) {
+            relations
+        } else {
+            return Ok(false);
+        };
+
+        // Extract the sender pattern from the condition
+        let sender_pattern = if let Some(sender) = sender {
+            Some(sender.as_ref())
+        } else if let Some(sender_type) = sender_type {
+            if sender_type == "user_id" {
+                if let Some(user_id) = user_id {
+                    Some(user_id)
+                } else {
+                    return Ok(false);
+                }
+            } else {
+                warn!("Unrecognized sender_type: {sender_type}");
+                return Ok(false);
+            }
+        } else {
+            None
+        };
+
+        let mut sender_compiled_pattern = if let Some(pattern) = sender_pattern {
+            Some(get_glob_matcher(pattern, GlobMatchType::Whole)?)
+        } else {
+            None
+        };
+
+        let mut type_compiled_pattern = if let Some(pattern) = event_type_pattern {
+            Some(get_glob_matcher(pattern, GlobMatchType::Whole)?)
+        } else {
+            None
+        };
+
+        for (relation_sender, event_type) in relations {
+            if let Some(pattern) = &mut sender_compiled_pattern {
+                if !pattern.is_match(relation_sender)? {
+                    continue;
+                }
+            }
+
+            if let Some(pattern) = &mut type_compiled_pattern {
+                if !pattern.is_match(event_type)? {
+                    continue;
+                }
+            }
+
+            return Ok(true);
+        }
+
+        Ok(false)
+    }
+
+    /// Evaluates a `event_match` condition.
+    fn match_event_match(
+        &self,
+        event_match: &EventMatchCondition,
+        user_id: Option<&str>,
+    ) -> Result<bool, Error> {
+        let pattern = if let Some(pattern) = &event_match.pattern {
+            pattern
+        } else if let Some(pattern_type) = &event_match.pattern_type {
+            // The `pattern_type` can either be "user_id" or "user_localpart",
+            // either way if we don't have a `user_id` then the condition can't
+            // match.
+            let user_id = if let Some(user_id) = user_id {
+                user_id
+            } else {
+                return Ok(false);
+            };
+
+            match &**pattern_type {
+                "user_id" => user_id,
+                "user_localpart" => get_localpart_from_id(user_id)?,
+                _ => return Ok(false),
+            }
+        } else {
+            return Ok(false);
+        };
+
+        let haystack = if let Some(haystack) = self.flattened_keys.get(&*event_match.key) {
+            haystack
+        } else {
+            return Ok(false);
+        };
+
+        // For the content.body we match against "words", but for everything
+        // else we match against the entire value.
+        let match_type = if event_match.key == "content.body" {
+            GlobMatchType::Word
+        } else {
+            GlobMatchType::Whole
+        };
+
+        let mut compiled_pattern = get_glob_matcher(pattern, match_type)?;
+        compiled_pattern.is_match(haystack)
+    }
+
+    /// Match the member count against an 'is' condition
+    /// The `is` condition can be things like '>2', '==3' or even just '4'.
+    fn match_member_count(&self, is: &str) -> Result<bool, Error> {
+        let captures = INEQUALITY_EXPR.captures(is).context("bad 'is' clause")?;
+        let ineq = captures.get(1).map_or("==", |m| m.as_str());
+        let rhs: u64 = captures
+            .get(2)
+            .context("missing number")?
+            .as_str()
+            .parse()?;
+
+        let matches = match ineq {
+            "" | "==" => self.room_member_count == rhs,
+            "<" => self.room_member_count < rhs,
+            ">" => self.room_member_count > rhs,
+            ">=" => self.room_member_count >= rhs,
+            "<=" => self.room_member_count <= rhs,
+            _ => false,
+        };
+
+        Ok(matches)
+    }
+}
+
+#[test]
+fn push_rule_evaluator() {
+    let mut flattened_keys = BTreeMap::new();
+    flattened_keys.insert("content.body".to_string(), "foo bar bob hello".to_string());
+    let evaluator = PushRuleEvaluator::py_new(
+        flattened_keys,
+        10,
+        Some(0),
+        BTreeMap::new(),
+        BTreeMap::new(),
+        true,
+    )
+    .unwrap();
+
+    let result = evaluator.run(&FilteredPushRules::default(), None, Some("bob"));
+    assert_eq!(result.len(), 3);
+}
diff --git a/rust/src/push/mod.rs b/rust/src/push/mod.rs
index de6764e7c5..30fffc31ad 100644
--- a/rust/src/push/mod.rs
+++ b/rust/src/push/mod.rs
@@ -42,7 +42,6 @@
 //!
 //! The set of "base rules" are the list of rules that every user has by default. A
 //! user can modify their copy of the push rules in one of three ways:
-//!
 //!     1. Adding a new push rule of a certain kind
 //!     2. Changing the actions of a base rule
 //!     3. Enabling/disabling a base rule.
@@ -58,12 +57,16 @@ use std::collections::{BTreeMap, HashMap, HashSet};
 use anyhow::{Context, Error};
 use log::warn;
 use pyo3::prelude::*;
-use pythonize::pythonize;
+use pythonize::{depythonize, pythonize};
 use serde::de::Error as _;
 use serde::{Deserialize, Serialize};
 use serde_json::Value;
 
+use self::evaluator::PushRuleEvaluator;
+
 mod base_rules;
+pub mod evaluator;
+pub mod utils;
 
 /// Called when registering modules with python.
 pub fn register_module(py: Python<'_>, m: &PyModule) -> PyResult<()> {
@@ -71,6 +74,7 @@ pub fn register_module(py: Python<'_>, m: &PyModule) -> PyResult<()> {
     child_module.add_class::<PushRule>()?;
     child_module.add_class::<PushRules>()?;
     child_module.add_class::<FilteredPushRules>()?;
+    child_module.add_class::<PushRuleEvaluator>()?;
     child_module.add_function(wrap_pyfunction!(get_base_rule_ids, m)?)?;
 
     m.add_submodule(child_module)?;
@@ -274,6 +278,8 @@ pub enum KnownCondition {
     #[serde(rename = "org.matrix.msc3772.relation_match")]
     RelationMatch {
         rel_type: Cow<'static, str>,
+        #[serde(skip_serializing_if = "Option::is_none", rename = "type")]
+        event_type_pattern: Option<Cow<'static, str>>,
         #[serde(skip_serializing_if = "Option::is_none")]
         sender: Option<Cow<'static, str>>,
         #[serde(skip_serializing_if = "Option::is_none")]
@@ -287,20 +293,26 @@ impl IntoPy<PyObject> for Condition {
     }
 }
 
+impl<'source> FromPyObject<'source> for Condition {
+    fn extract(ob: &'source PyAny) -> PyResult<Self> {
+        Ok(depythonize(ob)?)
+    }
+}
+
 /// The body of a [`Condition::EventMatch`]
 #[derive(Serialize, Deserialize, Debug, Clone)]
 pub struct EventMatchCondition {
-    key: Cow<'static, str>,
+    pub key: Cow<'static, str>,
     #[serde(skip_serializing_if = "Option::is_none")]
-    pattern: Option<Cow<'static, str>>,
+    pub pattern: Option<Cow<'static, str>>,
     #[serde(skip_serializing_if = "Option::is_none")]
-    pattern_type: Option<Cow<'static, str>>,
+    pub pattern_type: Option<Cow<'static, str>>,
 }
 
 /// The collection of push rules for a user.
 #[derive(Debug, Clone, Default)]
 #[pyclass(frozen)]
-struct PushRules {
+pub struct PushRules {
     /// Custom push rules that override a base rule.
     overridden_base_rules: HashMap<Cow<'static, str>, PushRule>,
 
@@ -319,7 +331,7 @@ struct PushRules {
 #[pymethods]
 impl PushRules {
     #[new]
-    fn new(rules: Vec<PushRule>) -> PushRules {
+    pub fn new(rules: Vec<PushRule>) -> PushRules {
         let mut push_rules: PushRules = Default::default();
 
         for rule in rules {
@@ -396,7 +408,7 @@ pub struct FilteredPushRules {
 #[pymethods]
 impl FilteredPushRules {
     #[new]
-    fn py_new(
+    pub fn py_new(
         push_rules: PushRules,
         enabled_map: BTreeMap<String, bool>,
         msc3786_enabled: bool,
diff --git a/rust/src/push/utils.rs b/rust/src/push/utils.rs
new file mode 100644
index 0000000000..8759340473
--- /dev/null
+++ b/rust/src/push/utils.rs
@@ -0,0 +1,215 @@
+// Copyright 2022 The Matrix.org Foundation C.I.C.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+use anyhow::bail;
+use anyhow::Context;
+use anyhow::Error;
+use lazy_static::lazy_static;
+use regex;
+use regex::Regex;
+use regex::RegexBuilder;
+
+lazy_static! {
+    /// Matches runs of non-wildcard characters followed by wildcard characters.
+    static ref WILDCARD_RUN: Regex = Regex::new(r"([^\?\*]*)([\?\*]*)").expect("valid regex");
+}
+
+/// Extract the localpart from a Matrix style ID
+pub(crate) fn get_localpart_from_id(id: &str) -> Result<&str, Error> {
+    let (localpart, _) = id
+        .split_once(':')
+        .with_context(|| format!("ID does not contain colon: {id}"))?;
+
+    // We need to strip off the first character, which is the ID type.
+    if localpart.is_empty() {
+        bail!("Invalid ID {id}");
+    }
+
+    Ok(&localpart[1..])
+}
+
+/// Used by `glob_to_regex` to specify what to match the regex against.
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub enum GlobMatchType {
+    /// The generated regex will match against the entire input.
+    Whole,
+    /// The generated regex will match against words.
+    Word,
+}
+
+/// Convert a "glob" style expression to a regex, anchoring either to the entire
+/// input or to individual words.
+pub fn glob_to_regex(glob: &str, match_type: GlobMatchType) -> Result<Regex, Error> {
+    let mut chunks = Vec::new();
+
+    // Patterns with wildcards must be simplified to avoid performance cliffs
+    // - The glob `?**?**?` is equivalent to the glob `???*`
+    // - The glob `???*` is equivalent to the regex `.{3,}`
+    for captures in WILDCARD_RUN.captures_iter(glob) {
+        if let Some(chunk) = captures.get(1) {
+            chunks.push(regex::escape(chunk.as_str()));
+        }
+
+        if let Some(wildcards) = captures.get(2) {
+            if wildcards.as_str() == "" {
+                continue;
+            }
+
+            let question_marks = wildcards.as_str().chars().filter(|c| *c == '?').count();
+
+            if wildcards.as_str().contains('*') {
+                chunks.push(format!(".{{{question_marks},}}"));
+            } else {
+                chunks.push(format!(".{{{question_marks}}}"));
+            }
+        }
+    }
+
+    let joined = chunks.join("");
+
+    let regex_str = match match_type {
+        GlobMatchType::Whole => format!(r"\A{joined}\z"),
+
+        // `^|\W` and `\W|$` handle the case where `pattern` starts or ends with a non-word
+        // character.
+        GlobMatchType::Word => format!(r"(?:^|\b|\W){joined}(?:\b|\W|$)"),
+    };
+
+    Ok(RegexBuilder::new(&regex_str)
+        .case_insensitive(true)
+        .build()?)
+}
+
+/// Compiles the glob into a `Matcher`.
+pub fn get_glob_matcher(glob: &str, match_type: GlobMatchType) -> Result<Matcher, Error> {
+    // There are a number of shortcuts we can make if the glob doesn't contain a
+    // wild card.
+    let matcher = if glob.contains(['*', '?']) {
+        let regex = glob_to_regex(glob, match_type)?;
+        Matcher::Regex(regex)
+    } else if match_type == GlobMatchType::Whole {
+        // If there aren't any wildcards and we're matching the whole thing,
+        // then we simply can do a case-insensitive string match.
+        Matcher::Whole(glob.to_lowercase())
+    } else {
+        // Otherwise, if we're matching against words then can first check
+        // if the haystack contains the glob at all.
+        Matcher::Word {
+            word: glob.to_lowercase(),
+            regex: None,
+        }
+    };
+
+    Ok(matcher)
+}
+
+/// Matches against a glob
+pub enum Matcher {
+    /// Plain regex matching.
+    Regex(Regex),
+
+    /// Case-insensitive equality.
+    Whole(String),
+
+    /// Word matching. `regex` is a cache of calling [`glob_to_regex`] on word.
+    Word { word: String, regex: Option<Regex> },
+}
+
+impl Matcher {
+    /// Checks if the glob matches the given haystack.
+    pub fn is_match(&mut self, haystack: &str) -> Result<bool, Error> {
+        // We want to to do case-insensitive matching, so we convert to
+        // lowercase first.
+        let haystack = haystack.to_lowercase();
+
+        match self {
+            Matcher::Regex(regex) => Ok(regex.is_match(&haystack)),
+            Matcher::Whole(whole) => Ok(whole == &haystack),
+            Matcher::Word { word, regex } => {
+                // If we're looking for a literal word, then we first check if
+                // the haystack contains the word as a substring.
+                if !haystack.contains(&*word) {
+                    return Ok(false);
+                }
+
+                // If it does contain the word as a substring, then we need to
+                // check if it is an actual word by testing it against the regex.
+                let regex = if let Some(regex) = regex {
+                    regex
+                } else {
+                    let compiled_regex = glob_to_regex(word, GlobMatchType::Word)?;
+                    regex.insert(compiled_regex)
+                };
+
+                Ok(regex.is_match(&haystack))
+            }
+        }
+    }
+}
+
+#[test]
+fn test_get_domain_from_id() {
+    get_localpart_from_id("").unwrap_err();
+    get_localpart_from_id(":").unwrap_err();
+    get_localpart_from_id(":asd").unwrap_err();
+    get_localpart_from_id("::as::asad").unwrap_err();
+
+    assert_eq!(get_localpart_from_id("@test:foo").unwrap(), "test");
+    assert_eq!(get_localpart_from_id("@:").unwrap(), "");
+    assert_eq!(get_localpart_from_id("@test:foo:907").unwrap(), "test");
+}
+
+#[test]
+fn tset_glob() -> Result<(), Error> {
+    assert_eq!(
+        glob_to_regex("simple", GlobMatchType::Whole)?.as_str(),
+        r"\Asimple\z"
+    );
+    assert_eq!(
+        glob_to_regex("simple*", GlobMatchType::Whole)?.as_str(),
+        r"\Asimple.{0,}\z"
+    );
+    assert_eq!(
+        glob_to_regex("simple?", GlobMatchType::Whole)?.as_str(),
+        r"\Asimple.{1}\z"
+    );
+    assert_eq!(
+        glob_to_regex("simple?*?*", GlobMatchType::Whole)?.as_str(),
+        r"\Asimple.{2,}\z"
+    );
+    assert_eq!(
+        glob_to_regex("simple???", GlobMatchType::Whole)?.as_str(),
+        r"\Asimple.{3}\z"
+    );
+
+    assert_eq!(
+        glob_to_regex("escape.", GlobMatchType::Whole)?.as_str(),
+        r"\Aescape\.\z"
+    );
+
+    assert!(glob_to_regex("simple", GlobMatchType::Whole)?.is_match("simple"));
+    assert!(!glob_to_regex("simple", GlobMatchType::Whole)?.is_match("simples"));
+    assert!(glob_to_regex("simple*", GlobMatchType::Whole)?.is_match("simples"));
+    assert!(glob_to_regex("simple?", GlobMatchType::Whole)?.is_match("simples"));
+    assert!(glob_to_regex("simple*", GlobMatchType::Whole)?.is_match("simple"));
+
+    assert!(glob_to_regex("simple", GlobMatchType::Word)?.is_match("some simple."));
+    assert!(glob_to_regex("simple", GlobMatchType::Word)?.is_match("simple"));
+    assert!(!glob_to_regex("simple", GlobMatchType::Word)?.is_match("simples"));
+
+    assert!(glob_to_regex("@user:foo", GlobMatchType::Word)?.is_match("Some @user:foo test"));
+    assert!(glob_to_regex("@user:foo", GlobMatchType::Word)?.is_match("@user:foo"));
+
+    Ok(())
+}
diff --git a/scripts-dev/check_pydantic_models.py b/scripts-dev/check_pydantic_models.py
index d0fb811bdb..9f2b7ded5b 100755
--- a/scripts-dev/check_pydantic_models.py
+++ b/scripts-dev/check_pydantic_models.py
@@ -88,10 +88,9 @@ def make_wrapper(factory: Callable[P, R]) -> Callable[P, R]:
 
     @functools.wraps(factory)
     def wrapper(*args: P.args, **kwargs: P.kwargs) -> R:
-        # type-ignore: should be redundant once we can use https://github.com/python/mypy/pull/12668
-        if "strict" not in kwargs:  # type: ignore[attr-defined]
+        if "strict" not in kwargs:
             raise MissingStrictInConstrainedTypeException(factory.__name__)
-        if not kwargs["strict"]:  # type: ignore[index]
+        if not kwargs["strict"]:
             raise MissingStrictInConstrainedTypeException(factory.__name__)
         return factory(*args, **kwargs)
 
diff --git a/stubs/synapse/synapse_rust/push.pyi b/stubs/synapse/synapse_rust/push.pyi
index 93c4e69d42..fffb8419c6 100644
--- a/stubs/synapse/synapse_rust/push.pyi
+++ b/stubs/synapse/synapse_rust/push.pyi
@@ -1,4 +1,4 @@
-from typing import Any, Collection, Dict, Mapping, Sequence, Tuple, Union
+from typing import Any, Collection, Dict, Mapping, Optional, Sequence, Set, Tuple, Union
 
 from synapse.types import JsonDict
 
@@ -35,3 +35,20 @@ class FilteredPushRules:
     def rules(self) -> Collection[Tuple[PushRule, bool]]: ...
 
 def get_base_rule_ids() -> Collection[str]: ...
+
+class PushRuleEvaluator:
+    def __init__(
+        self,
+        flattened_keys: Mapping[str, str],
+        room_member_count: int,
+        sender_power_level: Optional[int],
+        notification_power_levels: Mapping[str, int],
+        relations: Mapping[str, Set[Tuple[str, str]]],
+        relation_match_enabled: bool,
+    ): ...
+    def run(
+        self,
+        push_rules: FilteredPushRules,
+        user_id: Optional[str],
+        display_name: Optional[str],
+    ) -> Collection[dict]: ...
diff --git a/synapse/_scripts/synapse_port_db.py b/synapse/_scripts/synapse_port_db.py
index 450ba462ba..5fa599e70e 100755
--- a/synapse/_scripts/synapse_port_db.py
+++ b/synapse/_scripts/synapse_port_db.py
@@ -107,7 +107,7 @@ BOOLEAN_COLUMNS = {
     "redactions": ["have_censored"],
     "room_stats_state": ["is_federatable"],
     "local_media_repository": ["safe_from_quarantine"],
-    "users": ["shadow_banned"],
+    "users": ["shadow_banned", "approved"],
     "e2e_fallback_keys_json": ["used"],
     "access_tokens": ["used"],
     "device_lists_changes_in_room": ["converted_to_destinations"],
diff --git a/synapse/api/constants.py b/synapse/api/constants.py
index 5747d36802..a4db0dd632 100644
--- a/synapse/api/constants.py
+++ b/synapse/api/constants.py
@@ -272,3 +272,14 @@ class PublicRoomsFilterFields:
 
     GENERIC_SEARCH_TERM: Final = "generic_search_term"
     ROOM_TYPES: Final = "room_types"
+
+
+class ApprovalNoticeMedium:
+    """Identifier for the medium this server will use to serve notice of approval for a
+    specific user's registration.
+
+    As defined in https://github.com/matrix-org/matrix-spec-proposals/blob/babolivier/m_not_approved/proposals/3866-user-not-approved-error.md
+    """
+
+    NONE = "org.matrix.msc3866.none"
+    EMAIL = "org.matrix.msc3866.email"
diff --git a/synapse/api/errors.py b/synapse/api/errors.py
index 1c6b53aa24..c606207569 100644
--- a/synapse/api/errors.py
+++ b/synapse/api/errors.py
@@ -106,6 +106,8 @@ class Codes(str, Enum):
     # Part of MSC3895.
     UNABLE_DUE_TO_PARTIAL_STATE = "ORG.MATRIX.MSC3895_UNABLE_DUE_TO_PARTIAL_STATE"
 
+    USER_AWAITING_APPROVAL = "ORG.MATRIX.MSC3866_USER_AWAITING_APPROVAL"
+
 
 class CodeMessageException(RuntimeError):
     """An exception with integer code and message string attributes.
@@ -566,6 +568,20 @@ class UnredactedContentDeletedError(SynapseError):
         return cs_error(self.msg, self.errcode, **extra)
 
 
+class NotApprovedError(SynapseError):
+    def __init__(
+        self,
+        msg: str,
+        approval_notice_medium: str,
+    ):
+        super().__init__(
+            code=403,
+            msg=msg,
+            errcode=Codes.USER_AWAITING_APPROVAL,
+            additional_fields={"approval_notice_medium": approval_notice_medium},
+        )
+
+
 def cs_error(msg: str, code: str = Codes.UNKNOWN, **kwargs: Any) -> "JsonDict":
     """Utility method for constructing an error response for client-server
     interactions.
diff --git a/synapse/app/_base.py b/synapse/app/_base.py
index 65f7dc81c4..62626c2fe7 100644
--- a/synapse/app/_base.py
+++ b/synapse/app/_base.py
@@ -98,9 +98,7 @@ def register_sighup(func: Callable[P, None], *args: P.args, **kwargs: P.kwargs)
         func: Function to be called when sent a SIGHUP signal.
         *args, **kwargs: args and kwargs to be passed to the target function.
     """
-    # This type-ignore should be redundant once we use a mypy release with
-    # https://github.com/python/mypy/pull/12668.
-    _sighup_callbacks.append((func, args, kwargs))  # type: ignore[arg-type]
+    _sighup_callbacks.append((func, args, kwargs))
 
 
 def start_worker_reactor(
diff --git a/synapse/app/admin_cmd.py b/synapse/app/admin_cmd.py
index 8a583d3ec6..3c8c00ea5b 100644
--- a/synapse/app/admin_cmd.py
+++ b/synapse/app/admin_cmd.py
@@ -53,9 +53,9 @@ logger = logging.getLogger("synapse.app.admin_cmd")
 
 class AdminCmdSlavedStore(
     SlavedFilteringStore,
-    SlavedDeviceStore,
     SlavedPushRuleStore,
     SlavedEventStore,
+    SlavedDeviceStore,
     TagsWorkerStore,
     DeviceInboxWorkerStore,
     AccountDataWorkerStore,
diff --git a/synapse/app/complement_fork_starter.py b/synapse/app/complement_fork_starter.py
index 89eb07df27..b22f315453 100644
--- a/synapse/app/complement_fork_starter.py
+++ b/synapse/app/complement_fork_starter.py
@@ -51,11 +51,18 @@ import argparse
 import importlib
 import itertools
 import multiprocessing
+import os
+import signal
 import sys
-from typing import Any, Callable, List
+from types import FrameType
+from typing import Any, Callable, List, Optional
 
 from twisted.internet.main import installReactor
 
+# a list of the original signal handlers, before we installed our custom ones.
+# We restore these in our child processes.
+_original_signal_handlers: dict[int, Any] = {}
+
 
 class ProxiedReactor:
     """
@@ -105,6 +112,11 @@ def _worker_entrypoint(
 
     sys.argv = args
 
+    # reset the custom signal handlers that we installed, so that the children start
+    # from a clean slate.
+    for sig, handler in _original_signal_handlers.items():
+        signal.signal(sig, handler)
+
     from twisted.internet.epollreactor import EPollReactor
 
     proxy_reactor._install_real_reactor(EPollReactor())
@@ -167,13 +179,29 @@ def main() -> None:
     update_proc.join()
     print("===== PREPARED DATABASE =====", file=sys.stderr)
 
+    processes: List[multiprocessing.Process] = []
+
+    # Install signal handlers to propagate signals to all our children, so that they
+    # shut down cleanly. This also inhibits our own exit, but that's good: we want to
+    # wait until the children have exited.
+    def handle_signal(signum: int, frame: Optional[FrameType]) -> None:
+        print(
+            f"complement_fork_starter: Caught signal {signum}. Stopping children.",
+            file=sys.stderr,
+        )
+        for p in processes:
+            if p.pid:
+                os.kill(p.pid, signum)
+
+    for sig in (signal.SIGINT, signal.SIGTERM):
+        _original_signal_handlers[sig] = signal.signal(sig, handle_signal)
+
     # At this point, we've imported all the main entrypoints for all the workers.
     # Now we basically just fork() out to create the workers we need.
     # Because we're using fork(), all the workers get a clone of this launcher's
     # memory space and don't need to repeat the work of loading the code!
     # Instead of using fork() directly, we use the multiprocessing library,
     # which uses fork() on Unix platforms.
-    processes = []
     for (func, worker_args) in zip(worker_functions, args_by_worker):
         process = multiprocessing.Process(
             target=_worker_entrypoint, args=(func, proxy_reactor, worker_args)
diff --git a/synapse/config/experimental.py b/synapse/config/experimental.py
index 933779c23a..31834fb27d 100644
--- a/synapse/config/experimental.py
+++ b/synapse/config/experimental.py
@@ -14,10 +14,25 @@
 
 from typing import Any
 
+import attr
+
 from synapse.config._base import Config
 from synapse.types import JsonDict
 
 
+@attr.s(auto_attribs=True, frozen=True, slots=True)
+class MSC3866Config:
+    """Configuration for MSC3866 (mandating approval for new users)"""
+
+    # Whether the base support for the approval process is enabled. This includes the
+    # ability for administrators to check and update the approval of users, even if no
+    # approval is currently required.
+    enabled: bool = False
+    # Whether to require that new users are approved by an admin before their account
+    # can be used. Note that this setting is ignored if 'enabled' is false.
+    require_approval_for_new_accounts: bool = False
+
+
 class ExperimentalConfig(Config):
     """Config section for enabling experimental features"""
 
@@ -97,6 +112,10 @@ class ExperimentalConfig(Config):
         # MSC3852: Expose last seen user agent field on /_matrix/client/v3/devices.
         self.msc3852_enabled: bool = experimental.get("msc3852_enabled", False)
 
+        # MSC3866: M_USER_AWAITING_APPROVAL error code
+        raw_msc3866_config = experimental.get("msc3866", {})
+        self.msc3866 = MSC3866Config(**raw_msc3866_config)
+
         # MSC3881: Remotely toggle push notifications for another client
         self.msc3881_enabled: bool = experimental.get("msc3881_enabled", False)
 
diff --git a/synapse/events/__init__.py b/synapse/events/__init__.py
index b2c9119fd0..030c3ca408 100644
--- a/synapse/events/__init__.py
+++ b/synapse/events/__init__.py
@@ -289,6 +289,10 @@ class _EventInternalMetadata:
         """
         return self._dict.get("historical", False)
 
+    def is_notifiable(self) -> bool:
+        """Whether this event can trigger a push notification"""
+        return not self.is_outlier() or self.is_out_of_band_membership()
+
 
 class EventBase(metaclass=abc.ABCMeta):
     @property
diff --git a/synapse/federation/sender/per_destination_queue.py b/synapse/federation/sender/per_destination_queue.py
index 72bc935452..baeea3f1cd 100644
--- a/synapse/federation/sender/per_destination_queue.py
+++ b/synapse/federation/sender/per_destination_queue.py
@@ -646,29 +646,32 @@ class _TransactionQueueManager:
 
         # We start by fetching device related EDUs, i.e device updates and to
         # device messages. We have to keep 2 free slots for presence and rr_edus.
-        limit = MAX_EDUS_PER_TRANSACTION - 2
-
-        device_update_edus, dev_list_id = await self.queue._get_device_update_edus(
-            limit
-        )
-
-        if device_update_edus:
-            self._device_list_id = dev_list_id
-        else:
-            self.queue._last_device_list_stream_id = dev_list_id
-
-        limit -= len(device_update_edus)
+        device_edu_limit = MAX_EDUS_PER_TRANSACTION - 2
 
+        # We prioritize to-device messages so that existing encryption channels
+        # work. We also keep a few slots spare (by reducing the limit) so that
+        # we can still trickle out some device list updates.
         (
             to_device_edus,
             device_stream_id,
-        ) = await self.queue._get_to_device_message_edus(limit)
+        ) = await self.queue._get_to_device_message_edus(device_edu_limit - 10)
 
         if to_device_edus:
             self._device_stream_id = device_stream_id
         else:
             self.queue._last_device_stream_id = device_stream_id
 
+        device_edu_limit -= len(to_device_edus)
+
+        device_update_edus, dev_list_id = await self.queue._get_device_update_edus(
+            device_edu_limit
+        )
+
+        if device_update_edus:
+            self._device_list_id = dev_list_id
+        else:
+            self.queue._last_device_list_stream_id = dev_list_id
+
         pending_edus = device_update_edus + to_device_edus
 
         # Now add the read receipt EDU.
diff --git a/synapse/handlers/admin.py b/synapse/handlers/admin.py
index cf9f19608a..f2989cc4a2 100644
--- a/synapse/handlers/admin.py
+++ b/synapse/handlers/admin.py
@@ -32,6 +32,7 @@ class AdminHandler:
         self.store = hs.get_datastores().main
         self._storage_controllers = hs.get_storage_controllers()
         self._state_storage_controller = self._storage_controllers.state
+        self._msc3866_enabled = hs.config.experimental.msc3866.enabled
 
     async def get_whois(self, user: UserID) -> JsonDict:
         connections = []
@@ -75,6 +76,10 @@ class AdminHandler:
             "is_guest",
         }
 
+        if self._msc3866_enabled:
+            # Only include the approved flag if support for MSC3866 is enabled.
+            user_info_to_return.add("approved")
+
         # Restrict returned keys to a known set.
         user_info_dict = {
             key: value
diff --git a/synapse/handlers/auth.py b/synapse/handlers/auth.py
index eacd631ee0..f5f0e0e7a7 100644
--- a/synapse/handlers/auth.py
+++ b/synapse/handlers/auth.py
@@ -1009,6 +1009,17 @@ class AuthHandler:
             return res[0]
         return None
 
+    async def is_user_approved(self, user_id: str) -> bool:
+        """Checks if a user is approved and therefore can be allowed to log in.
+
+        Args:
+            user_id: the user to check the approval status of.
+
+        Returns:
+            A boolean that is True if the user is approved, False otherwise.
+        """
+        return await self.store.is_user_approved(user_id)
+
     async def _find_user_id_and_pwd_hash(
         self, user_id: str
     ) -> Optional[Tuple[str, str]]:
diff --git a/synapse/handlers/device.py b/synapse/handlers/device.py
index 961f8eb186..a3b11613e9 100644
--- a/synapse/handlers/device.py
+++ b/synapse/handlers/device.py
@@ -273,11 +273,9 @@ class DeviceWorkerHandler:
             possibly_left = possibly_changed | possibly_left
 
             # Double check if we still share rooms with the given user.
-            users_rooms = await self.store.get_rooms_for_users_with_stream_ordering(
-                possibly_left
-            )
+            users_rooms = await self.store.get_rooms_for_users(possibly_left)
             for changed_user_id, entries in users_rooms.items():
-                if any(e.room_id in room_ids for e in entries):
+                if any(rid in room_ids for rid in entries):
                     possibly_left.discard(changed_user_id)
                 else:
                     possibly_joined.discard(changed_user_id)
@@ -309,6 +307,17 @@ class DeviceWorkerHandler:
             "self_signing_key": self_signing_key,
         }
 
+    async def handle_room_un_partial_stated(self, room_id: str) -> None:
+        """Handles sending appropriate device list updates in a room that has
+        gone from partial to full state.
+        """
+
+        # TODO(faster_joins): worker mode support
+        #   https://github.com/matrix-org/synapse/issues/12994
+        logger.error(
+            "Trying handling device list state for partial join: not supported on workers."
+        )
+
 
 class DeviceHandler(DeviceWorkerHandler):
     def __init__(self, hs: "HomeServer"):
@@ -746,6 +755,95 @@ class DeviceHandler(DeviceWorkerHandler):
         finally:
             self._handle_new_device_update_is_processing = False
 
+    async def handle_room_un_partial_stated(self, room_id: str) -> None:
+        """Handles sending appropriate device list updates in a room that has
+        gone from partial to full state.
+        """
+
+        # We defer to the device list updater to handle pending remote device
+        # list updates.
+        await self.device_list_updater.handle_room_un_partial_stated(room_id)
+
+        # Replay local updates.
+        (
+            join_event_id,
+            device_lists_stream_id,
+        ) = await self.store.get_join_event_id_and_device_lists_stream_id_for_partial_state(
+            room_id
+        )
+
+        # Get the local device list changes that have happened in the room since
+        # we started joining. If there are no updates there's nothing left to do.
+        changes = await self.store.get_device_list_changes_in_room(
+            room_id, device_lists_stream_id
+        )
+        local_changes = {(u, d) for u, d in changes if self.hs.is_mine_id(u)}
+        if not local_changes:
+            return
+
+        # Note: We have persisted the full state at this point, we just haven't
+        # cleared the `partial_room` flag.
+        join_state_ids = await self._state_storage.get_state_ids_for_event(
+            join_event_id, await_full_state=False
+        )
+        current_state_ids = await self.store.get_partial_current_state_ids(room_id)
+
+        # Now we need to work out all servers that might have been in the room
+        # at any point during our join.
+
+        # First we look for any membership states that have changed between the
+        # initial join and now...
+        all_keys = set(join_state_ids)
+        all_keys.update(current_state_ids)
+
+        potentially_changed_hosts = set()
+        for etype, state_key in all_keys:
+            if etype != EventTypes.Member:
+                continue
+
+            prev = join_state_ids.get((etype, state_key))
+            current = current_state_ids.get((etype, state_key))
+
+            if prev != current:
+                potentially_changed_hosts.add(get_domain_from_id(state_key))
+
+        # ... then we add all the hosts that are currently joined to the room...
+        current_hosts_in_room = await self.store.get_current_hosts_in_room(room_id)
+        potentially_changed_hosts.update(current_hosts_in_room)
+
+        # ... and finally we remove any hosts that we were told about, as we
+        # will have sent device list updates to those hosts when they happened.
+        known_hosts_at_join = await self.store.get_partial_state_servers_at_join(
+            room_id
+        )
+        potentially_changed_hosts.difference_update(known_hosts_at_join)
+
+        potentially_changed_hosts.discard(self.server_name)
+
+        if not potentially_changed_hosts:
+            # Nothing to do.
+            return
+
+        logger.info(
+            "Found %d changed hosts to send device list updates to",
+            len(potentially_changed_hosts),
+        )
+
+        for user_id, device_id in local_changes:
+            await self.store.add_device_list_outbound_pokes(
+                user_id=user_id,
+                device_id=device_id,
+                room_id=room_id,
+                stream_id=None,
+                hosts=potentially_changed_hosts,
+                context=None,
+            )
+
+        # Notify things that device lists need to be sent out.
+        self.notifier.notify_replication()
+        for host in potentially_changed_hosts:
+            self.federation_sender.send_device_messages(host, immediate=False)
+
 
 def _update_device_from_client_ips(
     device: JsonDict, client_ips: Mapping[Tuple[str, str], Mapping[str, Any]]
@@ -836,6 +934,16 @@ class DeviceListUpdater:
             )
             return
 
+        # Check if we are partially joining any rooms. If so we need to store
+        # all device list updates so that we can handle them correctly once we
+        # know who is in the room.
+        partial_rooms = await self.store.get_partial_state_rooms_and_servers()
+        if partial_rooms:
+            await self.store.add_remote_device_list_to_pending(
+                user_id,
+                device_id,
+            )
+
         room_ids = await self.store.get_rooms_for_user(user_id)
         if not room_ids:
             # We don't share any rooms with this user. Ignore update, as we
@@ -1175,3 +1283,35 @@ class DeviceListUpdater:
             device_ids.append(verify_key.version)
 
         return device_ids
+
+    async def handle_room_un_partial_stated(self, room_id: str) -> None:
+        """Handles sending appropriate device list updates in a room that has
+        gone from partial to full state.
+        """
+
+        pending_updates = (
+            await self.store.get_pending_remote_device_list_updates_for_room(room_id)
+        )
+
+        for user_id, device_id in pending_updates:
+            logger.info(
+                "Got pending device list update in room %s: %s / %s",
+                room_id,
+                user_id,
+                device_id,
+            )
+            position = await self.store.add_device_change_to_streams(
+                user_id,
+                [device_id],
+                room_ids=[room_id],
+            )
+
+            if not position:
+                # This should only happen if there are no updates, which
+                # shouldn't happen when we've passed in a non-empty set of
+                # device IDs.
+                continue
+
+            self.device_handler.notifier.on_new_event(
+                StreamKeyType.DEVICE_LIST, position, rooms=[room_id]
+            )
diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py
index 7b9d07e4f5..610432b55a 100644
--- a/synapse/handlers/federation.py
+++ b/synapse/handlers/federation.py
@@ -38,7 +38,7 @@ from signedjson.sign import verify_signed_json
 from unpaddedbase64 import decode_base64
 
 from synapse import event_auth
-from synapse.api.constants import EventContentFields, EventTypes, Membership
+from synapse.api.constants import MAX_DEPTH, EventContentFields, EventTypes, Membership
 from synapse.api.errors import (
     AuthError,
     CodeMessageException,
@@ -149,6 +149,8 @@ class FederationHandler:
         self.http_client = hs.get_proxied_blacklisted_http_client()
         self._replication = hs.get_replication_data_handler()
         self._federation_event_handler = hs.get_federation_event_handler()
+        self._device_handler = hs.get_device_handler()
+        self._bulk_push_rule_evaluator = hs.get_bulk_push_rule_evaluator()
 
         self._clean_room_for_join_client = ReplicationCleanRoomRestServlet.make_client(
             hs
@@ -209,7 +211,7 @@ class FederationHandler:
         current_depth: int,
         limit: int,
         *,
-        processing_start_time: int,
+        processing_start_time: Optional[int],
     ) -> bool:
         """
         Checks whether the `current_depth` is at or approaching any backfill
@@ -221,12 +223,23 @@ class FederationHandler:
             room_id: The room to backfill in.
             current_depth: The depth to check at for any upcoming backfill points.
             limit: The max number of events to request from the remote federated server.
-            processing_start_time: The time when `maybe_backfill` started
-                processing. Only used for timing.
+            processing_start_time: The time when `maybe_backfill` started processing.
+                Only used for timing. If `None`, no timing observation will be made.
         """
         backwards_extremities = [
             _BackfillPoint(event_id, depth, _BackfillPointType.BACKWARDS_EXTREMITY)
-            for event_id, depth in await self.store.get_backfill_points_in_room(room_id)
+            for event_id, depth in await self.store.get_backfill_points_in_room(
+                room_id=room_id,
+                current_depth=current_depth,
+                # We only need to end up with 5 extremities combined with the
+                # insertion event extremities to make the `/backfill` request
+                # but fetch an order of magnitude more to make sure there is
+                # enough even after we filter them by whether visible in the
+                # history. This isn't fool-proof as all backfill points within
+                # our limit could be filtered out but seems like a good amount
+                # to try with at least.
+                limit=50,
+            )
         ]
 
         insertion_events_to_be_backfilled: List[_BackfillPoint] = []
@@ -234,7 +247,12 @@ class FederationHandler:
             insertion_events_to_be_backfilled = [
                 _BackfillPoint(event_id, depth, _BackfillPointType.INSERTION_PONT)
                 for event_id, depth in await self.store.get_insertion_event_backward_extremities_in_room(
-                    room_id
+                    room_id=room_id,
+                    current_depth=current_depth,
+                    # We only need to end up with 5 extremities combined with
+                    # the backfill points to make the `/backfill` request ...
+                    # (see the other comment above for more context).
+                    limit=50,
                 )
             ]
         logger.debug(
@@ -243,10 +261,6 @@ class FederationHandler:
             insertion_events_to_be_backfilled,
         )
 
-        if not backwards_extremities and not insertion_events_to_be_backfilled:
-            logger.debug("Not backfilling as no extremeties found.")
-            return False
-
         # we now have a list of potential places to backpaginate from. We prefer to
         # start with the most recent (ie, max depth), so let's sort the list.
         sorted_backfill_points: List[_BackfillPoint] = sorted(
@@ -267,6 +281,33 @@ class FederationHandler:
             sorted_backfill_points,
         )
 
+        # If we have no backfill points lower than the `current_depth` then
+        # either we can a) bail or b) still attempt to backfill. We opt to try
+        # backfilling anyway just in case we do get relevant events.
+        if not sorted_backfill_points and current_depth != MAX_DEPTH:
+            logger.debug(
+                "_maybe_backfill_inner: all backfill points are *after* current depth. Trying again with later backfill points."
+            )
+            return await self._maybe_backfill_inner(
+                room_id=room_id,
+                # We use `MAX_DEPTH` so that we find all backfill points next
+                # time (all events are below the `MAX_DEPTH`)
+                current_depth=MAX_DEPTH,
+                limit=limit,
+                # We don't want to start another timing observation from this
+                # nested recursive call. The top-most call can record the time
+                # overall otherwise the smaller one will throw off the results.
+                processing_start_time=None,
+            )
+
+        # Even after recursing with `MAX_DEPTH`, we didn't find any
+        # backward extremities to backfill from.
+        if not sorted_backfill_points:
+            logger.debug(
+                "_maybe_backfill_inner: Not backfilling as no backward extremeties found."
+            )
+            return False
+
         # If we're approaching an extremity we trigger a backfill, otherwise we
         # no-op.
         #
@@ -276,47 +317,16 @@ class FederationHandler:
         # chose more than one times the limit in case of failure, but choosing a
         # much larger factor will result in triggering a backfill request much
         # earlier than necessary.
-        #
-        # XXX: shouldn't we do this *after* the filter by depth below? Again, we don't
-        # care about events that have happened after our current position.
-        #
-        max_depth = sorted_backfill_points[0].depth
-        if current_depth - 2 * limit > max_depth:
+        max_depth_of_backfill_points = sorted_backfill_points[0].depth
+        if current_depth - 2 * limit > max_depth_of_backfill_points:
             logger.debug(
                 "Not backfilling as we don't need to. %d < %d - 2 * %d",
-                max_depth,
+                max_depth_of_backfill_points,
                 current_depth,
                 limit,
             )
             return False
 
-        # We ignore extremities that have a greater depth than our current depth
-        # as:
-        #    1. we don't really care about getting events that have happened
-        #       after our current position; and
-        #    2. we have likely previously tried and failed to backfill from that
-        #       extremity, so to avoid getting "stuck" requesting the same
-        #       backfill repeatedly we drop those extremities.
-        #
-        # However, we need to check that the filtered extremities are non-empty.
-        # If they are empty then either we can a) bail or b) still attempt to
-        # backfill. We opt to try backfilling anyway just in case we do get
-        # relevant events.
-        #
-        filtered_sorted_backfill_points = [
-            t for t in sorted_backfill_points if t.depth <= current_depth
-        ]
-        if filtered_sorted_backfill_points:
-            logger.debug(
-                "_maybe_backfill_inner: backfill points before current depth: %s",
-                filtered_sorted_backfill_points,
-            )
-            sorted_backfill_points = filtered_sorted_backfill_points
-        else:
-            logger.debug(
-                "_maybe_backfill_inner: all backfill points are *after* current depth. Backfilling anyway."
-            )
-
         # For performance's sake, we only want to paginate from a particular extremity
         # if we can actually see the events we'll get. Otherwise, we'd just spend a lot
         # of resources to get redacted events. We check each extremity in turn and
@@ -402,11 +412,22 @@ class FederationHandler:
         # First we try hosts that are already in the room.
         # TODO: HEURISTIC ALERT.
         likely_domains = (
-            await self._storage_controllers.state.get_current_hosts_in_room(room_id)
+            await self._storage_controllers.state.get_current_hosts_in_room_ordered(
+                room_id
+            )
         )
 
         async def try_backfill(domains: Collection[str]) -> bool:
             # TODO: Should we try multiple of these at a time?
+
+            # Number of contacted remote homeservers that have denied our backfill
+            # request with a 4xx code.
+            denied_count = 0
+
+            # Maximum number of contacted remote homeservers that can deny our
+            # backfill request with 4xx codes before we give up.
+            max_denied_count = 5
+
             for dom in domains:
                 # We don't want to ask our own server for information we don't have
                 if dom == self.server_name:
@@ -425,13 +446,33 @@ class FederationHandler:
                     continue
                 except HttpResponseException as e:
                     if 400 <= e.code < 500:
-                        raise e.to_synapse_error()
+                        logger.warning(
+                            "Backfill denied from %s because %s [%d/%d]",
+                            dom,
+                            e,
+                            denied_count,
+                            max_denied_count,
+                        )
+                        denied_count += 1
+                        if denied_count >= max_denied_count:
+                            return False
+                        continue
 
                     logger.info("Failed to backfill from %s because %s", dom, e)
                     continue
                 except CodeMessageException as e:
                     if 400 <= e.code < 500:
-                        raise
+                        logger.warning(
+                            "Backfill denied from %s because %s [%d/%d]",
+                            dom,
+                            e,
+                            denied_count,
+                            max_denied_count,
+                        )
+                        denied_count += 1
+                        if denied_count >= max_denied_count:
+                            return False
+                        continue
 
                     logger.info("Failed to backfill from %s because %s", dom, e)
                     continue
@@ -450,10 +491,15 @@ class FederationHandler:
 
             return False
 
-        processing_end_time = self.clock.time_msec()
-        backfill_processing_before_timer.observe(
-            (processing_end_time - processing_start_time) / 1000
-        )
+        # If we have the `processing_start_time`, then we can make an
+        # observation. We wouldn't have the `processing_start_time` in the case
+        # where `_maybe_backfill_inner` is recursively called to find any
+        # backfill points regardless of `current_depth`.
+        if processing_start_time is not None:
+            processing_end_time = self.clock.time_msec()
+            backfill_processing_before_timer.observe(
+                (processing_end_time - processing_start_time) / 1000
+            )
 
         success = await try_backfill(likely_domains)
         if success:
@@ -581,7 +627,11 @@ class FederationHandler:
                 # Mark the room as having partial state.
                 # The background process is responsible for unmarking this flag,
                 # even if the join fails.
-                await self.store.store_partial_state_room(room_id, ret.servers_in_room)
+                await self.store.store_partial_state_room(
+                    room_id=room_id,
+                    servers=ret.servers_in_room,
+                    device_lists_stream_id=self.store.get_device_stream_token(),
+                )
 
             try:
                 max_stream_id = (
@@ -606,6 +656,14 @@ class FederationHandler:
                     room_id,
                 )
                 raise LimitExceededError(msg=e.msg, errcode=e.errcode, retry_after_ms=0)
+            else:
+                # Record the join event id for future use (when we finish the full
+                # join). We have to do this after persisting the event to keep foreign
+                # key constraints intact.
+                if ret.partial_state:
+                    await self.store.write_partial_state_rooms_join_event_id(
+                        room_id, event.event_id
+                    )
             finally:
                 # Always kick off the background process that asynchronously fetches
                 # state for the room.
@@ -944,9 +1002,15 @@ class FederationHandler:
         )
 
         context = EventContext.for_outlier(self._storage_controllers)
-        await self._federation_event_handler.persist_events_and_notify(
-            event.room_id, [(event, context)]
-        )
+
+        await self._bulk_push_rule_evaluator.action_for_event_by_user(event, context)
+        try:
+            await self._federation_event_handler.persist_events_and_notify(
+                event.room_id, [(event, context)]
+            )
+        except Exception:
+            await self.store.remove_push_actions_from_staging(event.event_id)
+            raise
 
         return event
 
@@ -1612,6 +1676,9 @@ class FederationHandler:
                 #   https://github.com/matrix-org/synapse/issues/12994
                 await self.state_handler.update_current_state(room_id)
 
+                logger.info("Handling any pending device list updates")
+                await self._device_handler.handle_room_un_partial_stated(room_id)
+
                 logger.info("Clearing partial-state flag for %s", room_id)
                 success = await self.store.clear_partial_state_room(room_id)
                 if success:
diff --git a/synapse/handlers/federation_event.py b/synapse/handlers/federation_event.py
index 0be1653168..aa0740b08d 100644
--- a/synapse/handlers/federation_event.py
+++ b/synapse/handlers/federation_event.py
@@ -2170,6 +2170,7 @@ class FederationEventHandler:
         if instance != self._instance_name:
             # Limit the number of events sent over replication. We choose 200
             # here as that is what we default to in `max_request_body_size(..)`
+            result = {}
             try:
                 for batch in batch_iter(event_and_contexts, 200):
                     result = await self._send_events(
diff --git a/synapse/handlers/message.py b/synapse/handlers/message.py
index 10b5dad030..5123c93893 100644
--- a/synapse/handlers/message.py
+++ b/synapse/handlers/message.py
@@ -56,13 +56,16 @@ from synapse.logging import tracing
 from synapse.logging.context import make_deferred_yieldable, run_in_background
 from synapse.metrics.background_process_metrics import run_as_background_process
 from synapse.replication.http.send_event import ReplicationSendEventRestServlet
+from synapse.replication.http.send_events import ReplicationSendEventsRestServlet
 from synapse.storage.databases.main.events import PartialStateConflictError
 from synapse.storage.databases.main.events_worker import EventRedactBehaviour
 from synapse.storage.state import StateFilter
 from synapse.types import (
     MutableStateMap,
+    PersistedEventPosition,
     Requester,
     RoomAlias,
+    StateMap,
     StreamToken,
     UserID,
     create_requester,
@@ -492,6 +495,7 @@ class EventCreationHandler:
             self.membership_types_to_include_profile_data_in.add(Membership.INVITE)
 
         self.send_event = ReplicationSendEventRestServlet.make_client(hs)
+        self.send_events = ReplicationSendEventsRestServlet.make_client(hs)
 
         self.request_ratelimiter = hs.get_request_ratelimiter()
 
@@ -567,9 +571,17 @@ class EventCreationHandler:
         outlier: bool = False,
         historical: bool = False,
         depth: Optional[int] = None,
+        state_map: Optional[StateMap[str]] = None,
+        for_batch: bool = False,
+        current_state_group: Optional[int] = None,
     ) -> Tuple[EventBase, EventContext]:
         """
-        Given a dict from a client, create a new event.
+        Given a dict from a client, create a new event. If bool for_batch is true, will
+        create an event using the prev_event_ids, and will create an event context for
+        the event using the parameters state_map and current_state_group, thus these parameters
+        must be provided in this case if for_batch is True. The subsequently created event
+        and context are suitable for being batched up and bulk persisted to the database
+        with other similarly created events.
 
         Creates an FrozenEvent object, filling out auth_events, prev_events,
         etc.
@@ -612,16 +624,27 @@ class EventCreationHandler:
             outlier: Indicates whether the event is an `outlier`, i.e. if
                 it's from an arbitrary point and floating in the DAG as
                 opposed to being inline with the current DAG.
+
             historical: Indicates whether the message is being inserted
                 back in time around some existing events. This is used to skip
                 a few checks and mark the event as backfilled.
+
             depth: Override the depth used to order the event in the DAG.
                 Should normally be set to None, which will cause the depth to be calculated
                 based on the prev_events.
 
+            state_map: A state map of previously created events, used only when creating events
+                for batch persisting
+
+            for_batch: whether the event is being created for batch persisting to the db
+
+            current_state_group: the current state group, used only for creating events for
+                batch persisting
+
         Raises:
             ResourceLimitError if server is blocked to some resource being
             exceeded
+
         Returns:
             Tuple of created event, Context
         """
@@ -693,6 +716,9 @@ class EventCreationHandler:
             auth_event_ids=auth_event_ids,
             state_event_ids=state_event_ids,
             depth=depth,
+            state_map=state_map,
+            for_batch=for_batch,
+            current_state_group=current_state_group,
         )
 
         # In an ideal world we wouldn't need the second part of this condition. However,
@@ -707,10 +733,14 @@ class EventCreationHandler:
             # federation as well as those created locally. As of room v3, aliases events
             # can be created by users that are not in the room, therefore we have to
             # tolerate them in event_auth.check().
-            prev_state_ids = await context.get_prev_state_ids(
-                StateFilter.from_types([(EventTypes.Member, None)])
-            )
-            prev_event_id = prev_state_ids.get((EventTypes.Member, event.sender))
+            if for_batch:
+                assert state_map is not None
+                prev_event_id = state_map.get((EventTypes.Member, event.sender))
+            else:
+                prev_state_ids = await context.get_prev_state_ids(
+                    StateFilter.from_types([(EventTypes.Member, None)])
+                )
+                prev_event_id = prev_state_ids.get((EventTypes.Member, event.sender))
             prev_event = (
                 await self.store.get_event(prev_event_id, allow_none=True)
                 if prev_event_id
@@ -989,8 +1019,7 @@ class EventCreationHandler:
 
             ev = await self.handle_new_client_event(
                 requester=requester,
-                event=event,
-                context=context,
+                events_and_context=[(event, context)],
                 ratelimit=ratelimit,
                 ignore_shadow_ban=ignore_shadow_ban,
             )
@@ -1009,8 +1038,16 @@ class EventCreationHandler:
         auth_event_ids: Optional[List[str]] = None,
         state_event_ids: Optional[List[str]] = None,
         depth: Optional[int] = None,
+        state_map: Optional[StateMap[str]] = None,
+        for_batch: bool = False,
+        current_state_group: Optional[int] = None,
     ) -> Tuple[EventBase, EventContext]:
-        """Create a new event for a local client
+        """Create a new event for a local client. If bool for_batch is true, will
+        create an event using the prev_event_ids, and will create an event context for
+        the event using the parameters state_map and current_state_group, thus these parameters
+        must be provided in this case if for_batch is True. The subsequently created event
+        and context are suitable for being batched up and bulk persisted to the database
+        with other similarly created events.
 
         Args:
             builder:
@@ -1043,6 +1080,14 @@ class EventCreationHandler:
                 Should normally be set to None, which will cause the depth to be calculated
                 based on the prev_events.
 
+            state_map: A state map of previously created events, used only when creating events
+                for batch persisting
+
+            for_batch: whether the event is being created for batch persisting to the db
+
+            current_state_group: the current state group, used only for creating events for
+                batch persisting
+
         Returns:
             Tuple of created event, context
         """
@@ -1095,64 +1140,76 @@ class EventCreationHandler:
                 builder.type == EventTypes.Create or prev_event_ids
             ), "Attempting to create a non-m.room.create event with no prev_events"
 
-        event = await builder.build(
-            prev_event_ids=prev_event_ids,
-            auth_event_ids=auth_event_ids,
-            depth=depth,
-        )
+        if for_batch:
+            assert prev_event_ids is not None
+            assert state_map is not None
+            assert current_state_group is not None
+            auth_ids = self._event_auth_handler.compute_auth_events(builder, state_map)
+            event = await builder.build(
+                prev_event_ids=prev_event_ids, auth_event_ids=auth_ids, depth=depth
+            )
+            context = await self.state.compute_event_context_for_batched(
+                event, state_map, current_state_group
+            )
+        else:
+            event = await builder.build(
+                prev_event_ids=prev_event_ids,
+                auth_event_ids=auth_event_ids,
+                depth=depth,
+            )
 
-        # Pass on the outlier property from the builder to the event
-        # after it is created
-        if builder.internal_metadata.outlier:
-            event.internal_metadata.outlier = True
-            context = EventContext.for_outlier(self._storage_controllers)
-        elif (
-            event.type == EventTypes.MSC2716_INSERTION
-            and state_event_ids
-            and builder.internal_metadata.is_historical()
-        ):
-            # Add explicit state to the insertion event so it has state to derive
-            # from even though it's floating with no `prev_events`. The rest of
-            # the batch can derive from this state and state_group.
-            #
-            # TODO(faster_joins): figure out how this works, and make sure that the
-            #   old state is complete.
-            #   https://github.com/matrix-org/synapse/issues/13003
-            metadata = await self.store.get_metadata_for_events(state_event_ids)
-
-            state_map_for_event: MutableStateMap[str] = {}
-            for state_id in state_event_ids:
-                data = metadata.get(state_id)
-                if data is None:
-                    # We're trying to persist a new historical batch of events
-                    # with the given state, e.g. via
-                    # `RoomBatchSendEventRestServlet`. The state can be inferred
-                    # by Synapse or set directly by the client.
-                    #
-                    # Either way, we should have persisted all the state before
-                    # getting here.
-                    raise Exception(
-                        f"State event {state_id} not found in DB,"
-                        " Synapse should have persisted it before using it."
-                    )
+            # Pass on the outlier property from the builder to the event
+            # after it is created
+            if builder.internal_metadata.outlier:
+                event.internal_metadata.outlier = True
+                context = EventContext.for_outlier(self._storage_controllers)
+            elif (
+                event.type == EventTypes.MSC2716_INSERTION
+                and state_event_ids
+                and builder.internal_metadata.is_historical()
+            ):
+                # Add explicit state to the insertion event so it has state to derive
+                # from even though it's floating with no `prev_events`. The rest of
+                # the batch can derive from this state and state_group.
+                #
+                # TODO(faster_joins): figure out how this works, and make sure that the
+                #   old state is complete.
+                #   https://github.com/matrix-org/synapse/issues/13003
+                metadata = await self.store.get_metadata_for_events(state_event_ids)
+
+                state_map_for_event: MutableStateMap[str] = {}
+                for state_id in state_event_ids:
+                    data = metadata.get(state_id)
+                    if data is None:
+                        # We're trying to persist a new historical batch of events
+                        # with the given state, e.g. via
+                        # `RoomBatchSendEventRestServlet`. The state can be inferred
+                        # by Synapse or set directly by the client.
+                        #
+                        # Either way, we should have persisted all the state before
+                        # getting here.
+                        raise Exception(
+                            f"State event {state_id} not found in DB,"
+                            " Synapse should have persisted it before using it."
+                        )
 
-                if data.state_key is None:
-                    raise Exception(
-                        f"Trying to set non-state event {state_id} as state"
-                    )
+                    if data.state_key is None:
+                        raise Exception(
+                            f"Trying to set non-state event {state_id} as state"
+                        )
 
-                state_map_for_event[(data.event_type, data.state_key)] = state_id
+                    state_map_for_event[(data.event_type, data.state_key)] = state_id
 
-            context = await self.state.compute_event_context(
-                event,
-                state_ids_before_event=state_map_for_event,
-                # TODO(faster_joins): check how MSC2716 works and whether we can have
-                #   partial state here
-                #   https://github.com/matrix-org/synapse/issues/13003
-                partial_state=False,
-            )
-        else:
-            context = await self.state.compute_event_context(event)
+                context = await self.state.compute_event_context(
+                    event,
+                    state_ids_before_event=state_map_for_event,
+                    # TODO(faster_joins): check how MSC2716 works and whether we can have
+                    #   partial state here
+                    #   https://github.com/matrix-org/synapse/issues/13003
+                    partial_state=False,
+                )
+            else:
+                context = await self.state.compute_event_context(event)
 
         if requester:
             context.app_service = requester.app_service
@@ -1238,13 +1295,13 @@ class EventCreationHandler:
     async def handle_new_client_event(
         self,
         requester: Requester,
-        event: EventBase,
-        context: EventContext,
+        events_and_context: List[Tuple[EventBase, EventContext]],
         ratelimit: bool = True,
         extra_users: Optional[List[UserID]] = None,
         ignore_shadow_ban: bool = False,
     ) -> EventBase:
-        """Processes a new event.
+        """Processes new events. Please note that if batch persisting events, an error in
+        handling any one of these events will result in all of the events being dropped.
 
         This includes deduplicating, checking auth, persisting,
         notifying users, sending to remote servers, etc.
@@ -1254,8 +1311,7 @@ class EventCreationHandler:
 
         Args:
             requester
-            event
-            context
+            events_and_context: A list of one or more tuples of event, context to be persisted
             ratelimit
             extra_users: Any extra users to notify about event
 
@@ -1273,62 +1329,63 @@ class EventCreationHandler:
         """
         extra_users = extra_users or []
 
-        # we don't apply shadow-banning to membership events here. Invites are blocked
-        # higher up the stack, and we allow shadow-banned users to send join and leave
-        # events as normal.
-        if (
-            event.type != EventTypes.Member
-            and not ignore_shadow_ban
-            and requester.shadow_banned
-        ):
-            # We randomly sleep a bit just to annoy the requester.
-            await self.clock.sleep(random.randint(1, 10))
-            raise ShadowBanError()
+        for event, context in events_and_context:
+            # we don't apply shadow-banning to membership events here. Invites are blocked
+            # higher up the stack, and we allow shadow-banned users to send join and leave
+            # events as normal.
+            if (
+                event.type != EventTypes.Member
+                and not ignore_shadow_ban
+                and requester.shadow_banned
+            ):
+                # We randomly sleep a bit just to annoy the requester.
+                await self.clock.sleep(random.randint(1, 10))
+                raise ShadowBanError()
 
-        if event.is_state():
-            prev_event = await self.deduplicate_state_event(event, context)
-            if prev_event is not None:
-                logger.info(
-                    "Not bothering to persist state event %s duplicated by %s",
-                    event.event_id,
-                    prev_event.event_id,
-                )
-                return prev_event
+            if event.is_state():
+                prev_event = await self.deduplicate_state_event(event, context)
+                if prev_event is not None:
+                    logger.info(
+                        "Not bothering to persist state event %s duplicated by %s",
+                        event.event_id,
+                        prev_event.event_id,
+                    )
+                    return prev_event
 
-        if event.internal_metadata.is_out_of_band_membership():
-            # the only sort of out-of-band-membership events we expect to see here are
-            # invite rejections and rescinded knocks that we have generated ourselves.
-            assert event.type == EventTypes.Member
-            assert event.content["membership"] == Membership.LEAVE
-        else:
-            try:
-                validate_event_for_room_version(event)
-                await self._event_auth_handler.check_auth_rules_from_context(
-                    event, context
-                )
-            except AuthError as err:
-                logger.warning("Denying new event %r because %s", event, err)
-                raise err
+            if event.internal_metadata.is_out_of_band_membership():
+                # the only sort of out-of-band-membership events we expect to see here are
+                # invite rejections and rescinded knocks that we have generated ourselves.
+                assert event.type == EventTypes.Member
+                assert event.content["membership"] == Membership.LEAVE
+            else:
+                try:
+                    validate_event_for_room_version(event)
+                    await self._event_auth_handler.check_auth_rules_from_context(
+                        event, context
+                    )
+                except AuthError as err:
+                    logger.warning("Denying new event %r because %s", event, err)
+                    raise err
 
-        # Ensure that we can round trip before trying to persist in db
-        try:
-            dump = json_encoder.encode(event.content)
-            json_decoder.decode(dump)
-        except Exception:
-            logger.exception("Failed to encode content: %r", event.content)
-            raise
+            # Ensure that we can round trip before trying to persist in db
+            try:
+                dump = json_encoder.encode(event.content)
+                json_decoder.decode(dump)
+            except Exception:
+                logger.exception("Failed to encode content: %r", event.content)
+                raise
 
         # We now persist the event (and update the cache in parallel, since we
         # don't want to block on it).
+        event, context = events_and_context[0]
         try:
             result, _ = await make_deferred_yieldable(
                 gather_results(
                     (
                         run_in_background(
-                            self._persist_event,
+                            self._persist_events,
                             requester=requester,
-                            event=event,
-                            context=context,
+                            events_and_context=events_and_context,
                             ratelimit=ratelimit,
                             extra_users=extra_users,
                         ),
@@ -1352,45 +1409,47 @@ class EventCreationHandler:
 
         return result
 
-    async def _persist_event(
+    async def _persist_events(
         self,
         requester: Requester,
-        event: EventBase,
-        context: EventContext,
+        events_and_context: List[Tuple[EventBase, EventContext]],
         ratelimit: bool = True,
         extra_users: Optional[List[UserID]] = None,
     ) -> EventBase:
-        """Actually persists the event. Should only be called by
+        """Actually persists new events. Should only be called by
         `handle_new_client_event`, and see its docstring for documentation of
-        the arguments.
+        the arguments. Please note that if batch persisting events, an error in
+        handling any one of these events will result in all of the events being dropped.
 
         PartialStateConflictError: if attempting to persist a partial state event in
             a room that has been un-partial stated.
         """
 
-        # Skip push notification actions for historical messages
-        # because we don't want to notify people about old history back in time.
-        # The historical messages also do not have the proper `context.current_state_ids`
-        # and `state_groups` because they have `prev_events` that aren't persisted yet
-        # (historical messages persisted in reverse-chronological order).
-        if not event.internal_metadata.is_historical():
-            with tracing.start_active_span("calculate_push_actions"):
-                await self._bulk_push_rule_evaluator.action_for_event_by_user(
-                    event, context
-                )
+        for event, context in events_and_context:
+            # Skip push notification actions for historical messages
+            # because we don't want to notify people about old history back in time.
+            # The historical messages also do not have the proper `context.current_state_ids`
+            # and `state_groups` because they have `prev_events` that aren't persisted yet
+            # (historical messages persisted in reverse-chronological order).
+            if not event.internal_metadata.is_historical():
+                with tracing.start_active_span("calculate_push_actions"):
+                    await self._bulk_push_rule_evaluator.action_for_event_by_user(
+                        event, context
+                    )
 
         try:
             # If we're a worker we need to hit out to the master.
-            writer_instance = self._events_shard_config.get_instance(event.room_id)
+            first_event, _ = events_and_context[0]
+            writer_instance = self._events_shard_config.get_instance(
+                first_event.room_id
+            )
             if writer_instance != self._instance_name:
                 try:
-                    result = await self.send_event(
+                    result = await self.send_events(
                         instance_name=writer_instance,
-                        event_id=event.event_id,
+                        events_and_context=events_and_context,
                         store=self.store,
                         requester=requester,
-                        event=event,
-                        context=context,
                         ratelimit=ratelimit,
                         extra_users=extra_users,
                     )
@@ -1400,6 +1459,11 @@ class EventCreationHandler:
                     raise
                 stream_id = result["stream_id"]
                 event_id = result["event_id"]
+
+                # If we batch persisted events we return the last persisted event, otherwise
+                # we return the one event that was persisted
+                event, _ = events_and_context[-1]
+
                 if event_id != event.event_id:
                     # If we get a different event back then it means that its
                     # been de-duplicated, so we replace the given event with the
@@ -1412,15 +1476,19 @@ class EventCreationHandler:
                     event.internal_metadata.stream_ordering = stream_id
                 return event
 
-            event = await self.persist_and_notify_client_event(
-                requester, event, context, ratelimit=ratelimit, extra_users=extra_users
+            event = await self.persist_and_notify_client_events(
+                requester,
+                events_and_context,
+                ratelimit=ratelimit,
+                extra_users=extra_users,
             )
 
             return event
         except Exception:
-            # Ensure that we actually remove the entries in the push actions
-            # staging area, if we calculated them.
-            await self.store.remove_push_actions_from_staging(event.event_id)
+            for event, _ in events_and_context:
+                # Ensure that we actually remove the entries in the push actions
+                # staging area, if we calculated them.
+                await self.store.remove_push_actions_from_staging(event.event_id)
             raise
 
     async def cache_joined_hosts_for_event(
@@ -1514,23 +1582,26 @@ class EventCreationHandler:
                 Codes.BAD_ALIAS,
             )
 
-    async def persist_and_notify_client_event(
+    async def persist_and_notify_client_events(
         self,
         requester: Requester,
-        event: EventBase,
-        context: EventContext,
+        events_and_context: List[Tuple[EventBase, EventContext]],
         ratelimit: bool = True,
         extra_users: Optional[List[UserID]] = None,
     ) -> EventBase:
-        """Called when we have fully built the event, have already
-        calculated the push actions for the event, and checked auth.
+        """Called when we have fully built the events, have already
+        calculated the push actions for the events, and checked auth.
 
         This should only be run on the instance in charge of persisting events.
 
+        Please note that if batch persisting events, an error in
+        handling any one of these events will result in all of the events being dropped.
+
         Returns:
-            The persisted event. This may be different than the given event if
-            it was de-duplicated (e.g. because we had already persisted an
-            event with the same transaction ID.)
+            The persisted event, if one event is passed in, or the last event in the
+            list in the case of batch persisting. If only one event was persisted, the
+            returned event may be different than the given event if it was de-duplicated
+            (e.g. because we had already persisted an event with the same transaction ID.)
 
         Raises:
             PartialStateConflictError: if attempting to persist a partial state event in
@@ -1538,277 +1609,297 @@ class EventCreationHandler:
         """
         extra_users = extra_users or []
 
-        assert self._storage_controllers.persistence is not None
-        assert self._events_shard_config.should_handle(
-            self._instance_name, event.room_id
-        )
+        for event, context in events_and_context:
+            assert self._events_shard_config.should_handle(
+                self._instance_name, event.room_id
+            )
 
-        if ratelimit:
-            # We check if this is a room admin redacting an event so that we
-            # can apply different ratelimiting. We do this by simply checking
-            # it's not a self-redaction (to avoid having to look up whether the
-            # user is actually admin or not).
-            is_admin_redaction = False
-            if event.type == EventTypes.Redaction:
-                assert event.redacts is not None
+            if ratelimit:
+                # We check if this is a room admin redacting an event so that we
+                # can apply different ratelimiting. We do this by simply checking
+                # it's not a self-redaction (to avoid having to look up whether the
+                # user is actually admin or not).
+                is_admin_redaction = False
+                if event.type == EventTypes.Redaction:
+                    assert event.redacts is not None
+
+                    original_event = await self.store.get_event(
+                        event.redacts,
+                        redact_behaviour=EventRedactBehaviour.as_is,
+                        get_prev_content=False,
+                        allow_rejected=False,
+                        allow_none=True,
+                    )
 
-                original_event = await self.store.get_event(
-                    event.redacts,
-                    redact_behaviour=EventRedactBehaviour.as_is,
-                    get_prev_content=False,
-                    allow_rejected=False,
-                    allow_none=True,
+                    is_admin_redaction = bool(
+                        original_event and event.sender != original_event.sender
+                    )
+
+                await self.request_ratelimiter.ratelimit(
+                    requester, is_admin_redaction=is_admin_redaction
                 )
 
-                is_admin_redaction = bool(
-                    original_event and event.sender != original_event.sender
+            # run checks/actions on event based on type
+            if event.type == EventTypes.Member and event.membership == Membership.JOIN:
+                (
+                    current_membership,
+                    _,
+                ) = await self.store.get_local_current_membership_for_user_in_room(
+                    event.state_key, event.room_id
                 )
+                if current_membership != Membership.JOIN:
+                    self._notifier.notify_user_joined_room(
+                        event.event_id, event.room_id
+                    )
 
-            await self.request_ratelimiter.ratelimit(
-                requester, is_admin_redaction=is_admin_redaction
-            )
+            await self._maybe_kick_guest_users(event, context)
 
-        if event.type == EventTypes.Member and event.membership == Membership.JOIN:
-            (
-                current_membership,
-                _,
-            ) = await self.store.get_local_current_membership_for_user_in_room(
-                event.state_key, event.room_id
-            )
-            if current_membership != Membership.JOIN:
-                self._notifier.notify_user_joined_room(event.event_id, event.room_id)
+            if event.type == EventTypes.CanonicalAlias:
+                # Validate a newly added alias or newly added alt_aliases.
 
-        await self._maybe_kick_guest_users(event, context)
+                original_alias = None
+                original_alt_aliases: object = []
 
-        if event.type == EventTypes.CanonicalAlias:
-            # Validate a newly added alias or newly added alt_aliases.
+                original_event_id = event.unsigned.get("replaces_state")
+                if original_event_id:
+                    original_alias_event = await self.store.get_event(original_event_id)
 
-            original_alias = None
-            original_alt_aliases: object = []
+                    if original_alias_event:
+                        original_alias = original_alias_event.content.get("alias", None)
+                        original_alt_aliases = original_alias_event.content.get(
+                            "alt_aliases", []
+                        )
 
-            original_event_id = event.unsigned.get("replaces_state")
-            if original_event_id:
-                original_event = await self.store.get_event(original_event_id)
+                # Check the alias is currently valid (if it has changed).
+                room_alias_str = event.content.get("alias", None)
+                directory_handler = self.hs.get_directory_handler()
+                if room_alias_str and room_alias_str != original_alias:
+                    await self._validate_canonical_alias(
+                        directory_handler, room_alias_str, event.room_id
+                    )
 
-                if original_event:
-                    original_alias = original_event.content.get("alias", None)
-                    original_alt_aliases = original_event.content.get("alt_aliases", [])
-
-            # Check the alias is currently valid (if it has changed).
-            room_alias_str = event.content.get("alias", None)
-            directory_handler = self.hs.get_directory_handler()
-            if room_alias_str and room_alias_str != original_alias:
-                await self._validate_canonical_alias(
-                    directory_handler, room_alias_str, event.room_id
-                )
+                # Check that alt_aliases is the proper form.
+                alt_aliases = event.content.get("alt_aliases", [])
+                if not isinstance(alt_aliases, (list, tuple)):
+                    raise SynapseError(
+                        400,
+                        "The alt_aliases property must be a list.",
+                        Codes.INVALID_PARAM,
+                    )
 
-            # Check that alt_aliases is the proper form.
-            alt_aliases = event.content.get("alt_aliases", [])
-            if not isinstance(alt_aliases, (list, tuple)):
-                raise SynapseError(
-                    400, "The alt_aliases property must be a list.", Codes.INVALID_PARAM
-                )
+                # If the old version of alt_aliases is of an unknown form,
+                # completely replace it.
+                if not isinstance(original_alt_aliases, (list, tuple)):
+                    # TODO: check that the original_alt_aliases' entries are all strings
+                    original_alt_aliases = []
+
+                # Check that each alias is currently valid.
+                new_alt_aliases = set(alt_aliases) - set(original_alt_aliases)
+                if new_alt_aliases:
+                    for alias_str in new_alt_aliases:
+                        await self._validate_canonical_alias(
+                            directory_handler, alias_str, event.room_id
+                        )
 
-            # If the old version of alt_aliases is of an unknown form,
-            # completely replace it.
-            if not isinstance(original_alt_aliases, (list, tuple)):
-                # TODO: check that the original_alt_aliases' entries are all strings
-                original_alt_aliases = []
+            federation_handler = self.hs.get_federation_handler()
 
-            # Check that each alias is currently valid.
-            new_alt_aliases = set(alt_aliases) - set(original_alt_aliases)
-            if new_alt_aliases:
-                for alias_str in new_alt_aliases:
-                    await self._validate_canonical_alias(
-                        directory_handler, alias_str, event.room_id
+            if event.type == EventTypes.Member:
+                if event.content["membership"] == Membership.INVITE:
+                    event.unsigned[
+                        "invite_room_state"
+                    ] = await self.store.get_stripped_room_state_from_event_context(
+                        context,
+                        self.room_prejoin_state_types,
+                        membership_user_id=event.sender,
                     )
 
-        federation_handler = self.hs.get_federation_handler()
+                    invitee = UserID.from_string(event.state_key)
+                    if not self.hs.is_mine(invitee):
+                        # TODO: Can we add signature from remote server in a nicer
+                        # way? If we have been invited by a remote server, we need
+                        # to get them to sign the event.
 
-        if event.type == EventTypes.Member:
-            if event.content["membership"] == Membership.INVITE:
-                event.unsigned[
-                    "invite_room_state"
-                ] = await self.store.get_stripped_room_state_from_event_context(
-                    context,
-                    self.room_prejoin_state_types,
-                    membership_user_id=event.sender,
-                )
+                        returned_invite = await federation_handler.send_invite(
+                            invitee.domain, event
+                        )
+                        event.unsigned.pop("room_state", None)
 
-                invitee = UserID.from_string(event.state_key)
-                if not self.hs.is_mine(invitee):
-                    # TODO: Can we add signature from remote server in a nicer
-                    # way? If we have been invited by a remote server, we need
-                    # to get them to sign the event.
+                        # TODO: Make sure the signatures actually are correct.
+                        event.signatures.update(returned_invite.signatures)
 
-                    returned_invite = await federation_handler.send_invite(
-                        invitee.domain, event
+                if event.content["membership"] == Membership.KNOCK:
+                    event.unsigned[
+                        "knock_room_state"
+                    ] = await self.store.get_stripped_room_state_from_event_context(
+                        context,
+                        self.room_prejoin_state_types,
                     )
-                    event.unsigned.pop("room_state", None)
 
-                    # TODO: Make sure the signatures actually are correct.
-                    event.signatures.update(returned_invite.signatures)
+            if event.type == EventTypes.Redaction:
+                assert event.redacts is not None
 
-            if event.content["membership"] == Membership.KNOCK:
-                event.unsigned[
-                    "knock_room_state"
-                ] = await self.store.get_stripped_room_state_from_event_context(
-                    context,
-                    self.room_prejoin_state_types,
+                original_event = await self.store.get_event(
+                    event.redacts,
+                    redact_behaviour=EventRedactBehaviour.as_is,
+                    get_prev_content=False,
+                    allow_rejected=False,
+                    allow_none=True,
                 )
 
-        if event.type == EventTypes.Redaction:
-            assert event.redacts is not None
+                room_version = await self.store.get_room_version_id(event.room_id)
+                room_version_obj = KNOWN_ROOM_VERSIONS[room_version]
 
-            original_event = await self.store.get_event(
-                event.redacts,
-                redact_behaviour=EventRedactBehaviour.as_is,
-                get_prev_content=False,
-                allow_rejected=False,
-                allow_none=True,
-            )
+                # we can make some additional checks now if we have the original event.
+                if original_event:
+                    if original_event.type == EventTypes.Create:
+                        raise AuthError(403, "Redacting create events is not permitted")
 
-            room_version = await self.store.get_room_version_id(event.room_id)
-            room_version_obj = KNOWN_ROOM_VERSIONS[room_version]
-
-            # we can make some additional checks now if we have the original event.
-            if original_event:
-                if original_event.type == EventTypes.Create:
-                    raise AuthError(403, "Redacting create events is not permitted")
-
-                if original_event.room_id != event.room_id:
-                    raise SynapseError(400, "Cannot redact event from a different room")
-
-                if original_event.type == EventTypes.ServerACL:
-                    raise AuthError(403, "Redacting server ACL events is not permitted")
-
-                # Add a little safety stop-gap to prevent people from trying to
-                # redact MSC2716 related events when they're in a room version
-                # which does not support it yet. We allow people to use MSC2716
-                # events in existing room versions but only from the room
-                # creator since it does not require any changes to the auth
-                # rules and in effect, the redaction algorithm . In the
-                # supported room version, we add the `historical` power level to
-                # auth the MSC2716 related events and adjust the redaction
-                # algorthim to keep the `historical` field around (redacting an
-                # event should only strip fields which don't affect the
-                # structural protocol level).
-                is_msc2716_event = (
-                    original_event.type == EventTypes.MSC2716_INSERTION
-                    or original_event.type == EventTypes.MSC2716_BATCH
-                    or original_event.type == EventTypes.MSC2716_MARKER
-                )
-                if not room_version_obj.msc2716_historical and is_msc2716_event:
-                    raise AuthError(
-                        403,
-                        "Redacting MSC2716 events is not supported in this room version",
-                    )
+                    if original_event.room_id != event.room_id:
+                        raise SynapseError(
+                            400, "Cannot redact event from a different room"
+                        )
 
-            event_types = event_auth.auth_types_for_event(event.room_version, event)
-            prev_state_ids = await context.get_prev_state_ids(
-                StateFilter.from_types(event_types)
-            )
+                    if original_event.type == EventTypes.ServerACL:
+                        raise AuthError(
+                            403, "Redacting server ACL events is not permitted"
+                        )
 
-            auth_events_ids = self._event_auth_handler.compute_auth_events(
-                event, prev_state_ids, for_verification=True
-            )
-            auth_events_map = await self.store.get_events(auth_events_ids)
-            auth_events = {(e.type, e.state_key): e for e in auth_events_map.values()}
+                    # Add a little safety stop-gap to prevent people from trying to
+                    # redact MSC2716 related events when they're in a room version
+                    # which does not support it yet. We allow people to use MSC2716
+                    # events in existing room versions but only from the room
+                    # creator since it does not require any changes to the auth
+                    # rules and in effect, the redaction algorithm . In the
+                    # supported room version, we add the `historical` power level to
+                    # auth the MSC2716 related events and adjust the redaction
+                    # algorthim to keep the `historical` field around (redacting an
+                    # event should only strip fields which don't affect the
+                    # structural protocol level).
+                    is_msc2716_event = (
+                        original_event.type == EventTypes.MSC2716_INSERTION
+                        or original_event.type == EventTypes.MSC2716_BATCH
+                        or original_event.type == EventTypes.MSC2716_MARKER
+                    )
+                    if not room_version_obj.msc2716_historical and is_msc2716_event:
+                        raise AuthError(
+                            403,
+                            "Redacting MSC2716 events is not supported in this room version",
+                        )
 
-            if event_auth.check_redaction(
-                room_version_obj, event, auth_events=auth_events
-            ):
-                # this user doesn't have 'redact' rights, so we need to do some more
-                # checks on the original event. Let's start by checking the original
-                # event exists.
-                if not original_event:
-                    raise NotFoundError("Could not find event %s" % (event.redacts,))
-
-                if event.user_id != original_event.user_id:
-                    raise AuthError(403, "You don't have permission to redact events")
-
-                # all the checks are done.
-                event.internal_metadata.recheck_redaction = False
-
-        if event.type == EventTypes.Create:
-            prev_state_ids = await context.get_prev_state_ids()
-            if prev_state_ids:
-                raise AuthError(403, "Changing the room create event is forbidden")
-
-        if event.type == EventTypes.MSC2716_INSERTION:
-            room_version = await self.store.get_room_version_id(event.room_id)
-            room_version_obj = KNOWN_ROOM_VERSIONS[room_version]
-
-            create_event = await self.store.get_create_event_for_room(event.room_id)
-            room_creator = create_event.content.get(EventContentFields.ROOM_CREATOR)
-
-            # Only check an insertion event if the room version
-            # supports it or the event is from the room creator.
-            if room_version_obj.msc2716_historical or (
-                self.config.experimental.msc2716_enabled
-                and event.sender == room_creator
-            ):
-                next_batch_id = event.content.get(
-                    EventContentFields.MSC2716_NEXT_BATCH_ID
+                event_types = event_auth.auth_types_for_event(event.room_version, event)
+                prev_state_ids = await context.get_prev_state_ids(
+                    StateFilter.from_types(event_types)
+                )
+
+                auth_events_ids = self._event_auth_handler.compute_auth_events(
+                    event, prev_state_ids, for_verification=True
                 )
-                conflicting_insertion_event_id = None
-                if next_batch_id:
-                    conflicting_insertion_event_id = (
-                        await self.store.get_insertion_event_id_by_batch_id(
-                            event.room_id, next_batch_id
+                auth_events_map = await self.store.get_events(auth_events_ids)
+                auth_events = {
+                    (e.type, e.state_key): e for e in auth_events_map.values()
+                }
+
+                if event_auth.check_redaction(
+                    room_version_obj, event, auth_events=auth_events
+                ):
+                    # this user doesn't have 'redact' rights, so we need to do some more
+                    # checks on the original event. Let's start by checking the original
+                    # event exists.
+                    if not original_event:
+                        raise NotFoundError(
+                            "Could not find event %s" % (event.redacts,)
                         )
+
+                    if event.user_id != original_event.user_id:
+                        raise AuthError(
+                            403, "You don't have permission to redact events"
+                        )
+
+                    # all the checks are done.
+                    event.internal_metadata.recheck_redaction = False
+
+            if event.type == EventTypes.Create:
+                prev_state_ids = await context.get_prev_state_ids()
+                if prev_state_ids:
+                    raise AuthError(403, "Changing the room create event is forbidden")
+
+            if event.type == EventTypes.MSC2716_INSERTION:
+                room_version = await self.store.get_room_version_id(event.room_id)
+                room_version_obj = KNOWN_ROOM_VERSIONS[room_version]
+
+                create_event = await self.store.get_create_event_for_room(event.room_id)
+                room_creator = create_event.content.get(EventContentFields.ROOM_CREATOR)
+
+                # Only check an insertion event if the room version
+                # supports it or the event is from the room creator.
+                if room_version_obj.msc2716_historical or (
+                    self.config.experimental.msc2716_enabled
+                    and event.sender == room_creator
+                ):
+                    next_batch_id = event.content.get(
+                        EventContentFields.MSC2716_NEXT_BATCH_ID
                     )
-                if conflicting_insertion_event_id is not None:
-                    # The current insertion event that we're processing is invalid
-                    # because an insertion event already exists in the room with the
-                    # same next_batch_id. We can't allow multiple because the batch
-                    # pointing will get weird, e.g. we can't determine which insertion
-                    # event the batch event is pointing to.
-                    raise SynapseError(
-                        HTTPStatus.BAD_REQUEST,
-                        "Another insertion event already exists with the same next_batch_id",
-                        errcode=Codes.INVALID_PARAM,
-                    )
+                    conflicting_insertion_event_id = None
+                    if next_batch_id:
+                        conflicting_insertion_event_id = (
+                            await self.store.get_insertion_event_id_by_batch_id(
+                                event.room_id, next_batch_id
+                            )
+                        )
+                    if conflicting_insertion_event_id is not None:
+                        # The current insertion event that we're processing is invalid
+                        # because an insertion event already exists in the room with the
+                        # same next_batch_id. We can't allow multiple because the batch
+                        # pointing will get weird, e.g. we can't determine which insertion
+                        # event the batch event is pointing to.
+                        raise SynapseError(
+                            HTTPStatus.BAD_REQUEST,
+                            "Another insertion event already exists with the same next_batch_id",
+                            errcode=Codes.INVALID_PARAM,
+                        )
 
-        # Mark any `m.historical` messages as backfilled so they don't appear
-        # in `/sync` and have the proper decrementing `stream_ordering` as we import
-        backfilled = False
-        if event.internal_metadata.is_historical():
-            backfilled = True
+            # Mark any `m.historical` messages as backfilled so they don't appear
+            # in `/sync` and have the proper decrementing `stream_ordering` as we import
+            backfilled = False
+            if event.internal_metadata.is_historical():
+                backfilled = True
 
-        # Note that this returns the event that was persisted, which may not be
-        # the same as we passed in if it was deduplicated due transaction IDs.
+        assert self._storage_controllers.persistence is not None
         (
-            event,
-            event_pos,
+            persisted_events,
             max_stream_token,
-        ) = await self._storage_controllers.persistence.persist_event(
-            event, context=context, backfilled=backfilled
+        ) = await self._storage_controllers.persistence.persist_events(
+            events_and_context, backfilled=backfilled
         )
 
-        if self._ephemeral_events_enabled:
-            # If there's an expiry timestamp on the event, schedule its expiry.
-            self._message_handler.maybe_schedule_expiry(event)
+        for event in persisted_events:
+            if self._ephemeral_events_enabled:
+                # If there's an expiry timestamp on the event, schedule its expiry.
+                self._message_handler.maybe_schedule_expiry(event)
 
-        async def _notify() -> None:
-            try:
-                await self.notifier.on_new_room_event(
-                    event, event_pos, max_stream_token, extra_users=extra_users
-                )
-            except Exception:
-                logger.exception(
-                    "Error notifying about new room event %s",
-                    event.event_id,
-                )
+            stream_ordering = event.internal_metadata.stream_ordering
+            assert stream_ordering is not None
+            pos = PersistedEventPosition(self._instance_name, stream_ordering)
+
+            async def _notify() -> None:
+                try:
+                    await self.notifier.on_new_room_event(
+                        event, pos, max_stream_token, extra_users=extra_users
+                    )
+                except Exception:
+                    logger.exception(
+                        "Error notifying about new room event %s",
+                        event.event_id,
+                    )
 
-        run_in_background(_notify)
+            run_in_background(_notify)
 
-        if event.type == EventTypes.Message:
-            # We don't want to block sending messages on any presence code. This
-            # matters as sometimes presence code can take a while.
-            run_in_background(self._bump_active_time, requester.user)
+            if event.type == EventTypes.Message:
+                # We don't want to block sending messages on any presence code. This
+                # matters as sometimes presence code can take a while.
+                run_in_background(self._bump_active_time, requester.user)
 
-        return event
+        return persisted_events[-1]
 
     async def _maybe_kick_guest_users(
         self, event: EventBase, context: EventContext
@@ -1897,8 +1988,7 @@ class EventCreationHandler:
                 # shadow-banned user.
                 await self.handle_new_client_event(
                     requester,
-                    event,
-                    context,
+                    events_and_context=[(event, context)],
                     ratelimit=False,
                     ignore_shadow_ban=True,
                 )
diff --git a/synapse/handlers/register.py b/synapse/handlers/register.py
index cfcadb34db..ca1c7a1866 100644
--- a/synapse/handlers/register.py
+++ b/synapse/handlers/register.py
@@ -220,6 +220,7 @@ class RegistrationHandler:
         by_admin: bool = False,
         user_agent_ips: Optional[List[Tuple[str, str]]] = None,
         auth_provider_id: Optional[str] = None,
+        approved: bool = False,
     ) -> str:
         """Registers a new client on the server.
 
@@ -246,6 +247,8 @@ class RegistrationHandler:
             user_agent_ips: Tuples of user-agents and IP addresses used
                 during the registration process.
             auth_provider_id: The SSO IdP the user used, if any.
+            approved: True if the new user should be considered already
+                approved by an administrator.
         Returns:
             The registered user_id.
         Raises:
@@ -307,6 +310,7 @@ class RegistrationHandler:
                 user_type=user_type,
                 address=address,
                 shadow_banned=shadow_banned,
+                approved=approved,
             )
 
             profile = await self.store.get_profileinfo(localpart)
@@ -695,6 +699,7 @@ class RegistrationHandler:
         user_type: Optional[str] = None,
         address: Optional[str] = None,
         shadow_banned: bool = False,
+        approved: bool = False,
     ) -> None:
         """Register user in the datastore.
 
@@ -713,6 +718,7 @@ class RegistrationHandler:
                 api.constants.UserTypes, or None for a normal user.
             address: the IP address used to perform the registration.
             shadow_banned: Whether to shadow-ban the user
+            approved: Whether to mark the user as approved by an administrator
         """
         if self.hs.config.worker.worker_app:
             await self._register_client(
@@ -726,6 +732,7 @@ class RegistrationHandler:
                 user_type=user_type,
                 address=address,
                 shadow_banned=shadow_banned,
+                approved=approved,
             )
         else:
             await self.store.register_user(
@@ -738,6 +745,7 @@ class RegistrationHandler:
                 admin=admin,
                 user_type=user_type,
                 shadow_banned=shadow_banned,
+                approved=approved,
             )
 
             # Only call the account validity module(s) on the main process, to avoid
diff --git a/synapse/handlers/room.py b/synapse/handlers/room.py
index 33e9a87002..57ab05ad25 100644
--- a/synapse/handlers/room.py
+++ b/synapse/handlers/room.py
@@ -301,8 +301,7 @@ class RoomCreationHandler:
         # now send the tombstone
         await self.event_creation_handler.handle_new_client_event(
             requester=requester,
-            event=tombstone_event,
-            context=tombstone_context,
+            events_and_context=[(tombstone_event, tombstone_context)],
         )
 
         state_filter = StateFilter.from_types(
@@ -716,7 +715,7 @@ class RoomCreationHandler:
 
         if (
             self._server_notices_mxid is not None
-            and requester.user.to_string() == self._server_notices_mxid
+            and user_id == self._server_notices_mxid
         ):
             # allow the server notices mxid to create rooms
             is_requester_admin = True
@@ -1042,7 +1041,9 @@ class RoomCreationHandler:
         creator_join_profile: Optional[JsonDict] = None,
         ratelimit: bool = True,
     ) -> Tuple[int, str, int]:
-        """Sends the initial events into a new room.
+        """Sends the initial events into a new room. Sends the room creation, membership,
+        and power level events into the room sequentially, then creates and batches up the
+        rest of the events to persist as a batch to the DB.
 
         `power_level_content_override` doesn't apply when initial state has
         power level state event content.
@@ -1053,13 +1054,23 @@ class RoomCreationHandler:
         """
 
         creator_id = creator.user.to_string()
-
         event_keys = {"room_id": room_id, "sender": creator_id, "state_key": ""}
-
         depth = 1
+
+        # the last event sent/persisted to the db
         last_sent_event_id: Optional[str] = None
 
-        def create(etype: str, content: JsonDict, **kwargs: Any) -> JsonDict:
+        # the most recently created event
+        prev_event: List[str] = []
+        # a map of event types, state keys -> event_ids. We collect these mappings this as events are
+        # created (but not persisted to the db) to determine state for future created events
+        # (as this info can't be pulled from the db)
+        state_map: MutableStateMap[str] = {}
+        # current_state_group of last event created. Used for computing event context of
+        # events to be batched
+        current_state_group = None
+
+        def create_event_dict(etype: str, content: JsonDict, **kwargs: Any) -> JsonDict:
             e = {"type": etype, "content": content}
 
             e.update(event_keys)
@@ -1067,32 +1078,51 @@ class RoomCreationHandler:
 
             return e
 
-        async def send(etype: str, content: JsonDict, **kwargs: Any) -> int:
-            nonlocal last_sent_event_id
+        async def create_event(
+            etype: str,
+            content: JsonDict,
+            for_batch: bool,
+            **kwargs: Any,
+        ) -> Tuple[EventBase, synapse.events.snapshot.EventContext]:
             nonlocal depth
+            nonlocal prev_event
 
-            event = create(etype, content, **kwargs)
-            logger.debug("Sending %s in new room", etype)
-            # Allow these events to be sent even if the user is shadow-banned to
-            # allow the room creation to complete.
-            (
-                sent_event,
-                last_stream_id,
-            ) = await self.event_creation_handler.create_and_send_nonmember_event(
+            event_dict = create_event_dict(etype, content, **kwargs)
+
+            new_event, new_context = await self.event_creation_handler.create_event(
                 creator,
-                event,
+                event_dict,
+                prev_event_ids=prev_event,
+                depth=depth,
+                state_map=state_map,
+                for_batch=for_batch,
+                current_state_group=current_state_group,
+            )
+            depth += 1
+            prev_event = [new_event.event_id]
+            state_map[(new_event.type, new_event.state_key)] = new_event.event_id
+
+            return new_event, new_context
+
+        async def send(
+            event: EventBase,
+            context: synapse.events.snapshot.EventContext,
+            creator: Requester,
+        ) -> int:
+            nonlocal last_sent_event_id
+
+            ev = await self.event_creation_handler.handle_new_client_event(
+                requester=creator,
+                events_and_context=[(event, context)],
                 ratelimit=False,
                 ignore_shadow_ban=True,
-                # Note: we don't pass state_event_ids here because this triggers
-                # an additional query per event to look them up from the events table.
-                prev_event_ids=[last_sent_event_id] if last_sent_event_id else [],
-                depth=depth,
             )
 
-            last_sent_event_id = sent_event.event_id
-            depth += 1
+            last_sent_event_id = ev.event_id
 
-            return last_stream_id
+            # we know it was persisted, so must have a stream ordering
+            assert ev.internal_metadata.stream_ordering
+            return ev.internal_metadata.stream_ordering
 
         try:
             config = self._presets_dict[preset_config]
@@ -1102,9 +1132,13 @@ class RoomCreationHandler:
             )
 
         creation_content.update({"creator": creator_id})
-        await send(etype=EventTypes.Create, content=creation_content)
+        creation_event, creation_context = await create_event(
+            EventTypes.Create, creation_content, False
+        )
 
         logger.debug("Sending %s in new room", EventTypes.Member)
+        await send(creation_event, creation_context, creator)
+
         # Room create event must exist at this point
         assert last_sent_event_id is not None
         member_event_id, _ = await self.room_member_handler.update_membership(
@@ -1118,15 +1152,22 @@ class RoomCreationHandler:
             prev_event_ids=[last_sent_event_id],
             depth=depth,
         )
-        last_sent_event_id = member_event_id
+        prev_event = [member_event_id]
+
+        # update the depth and state map here as the membership event has been created
+        # through a different code path
+        depth += 1
+        state_map[(EventTypes.Member, creator.user.to_string())] = member_event_id
 
         # We treat the power levels override specially as this needs to be one
         # of the first events that get sent into a room.
         pl_content = initial_state.pop((EventTypes.PowerLevels, ""), None)
         if pl_content is not None:
-            last_sent_stream_id = await send(
-                etype=EventTypes.PowerLevels, content=pl_content
+            power_event, power_context = await create_event(
+                EventTypes.PowerLevels, pl_content, False
             )
+            current_state_group = power_context._state_group
+            await send(power_event, power_context, creator)
         else:
             power_level_content: JsonDict = {
                 "users": {creator_id: 100},
@@ -1169,48 +1210,71 @@ class RoomCreationHandler:
             # apply those.
             if power_level_content_override:
                 power_level_content.update(power_level_content_override)
-
-            last_sent_stream_id = await send(
-                etype=EventTypes.PowerLevels, content=power_level_content
+            pl_event, pl_context = await create_event(
+                EventTypes.PowerLevels,
+                power_level_content,
+                False,
             )
+            current_state_group = pl_context._state_group
+            await send(pl_event, pl_context, creator)
 
+        events_to_send = []
         if room_alias and (EventTypes.CanonicalAlias, "") not in initial_state:
-            last_sent_stream_id = await send(
-                etype=EventTypes.CanonicalAlias,
-                content={"alias": room_alias.to_string()},
+            room_alias_event, room_alias_context = await create_event(
+                EventTypes.CanonicalAlias, {"alias": room_alias.to_string()}, True
             )
+            current_state_group = room_alias_context._state_group
+            events_to_send.append((room_alias_event, room_alias_context))
 
         if (EventTypes.JoinRules, "") not in initial_state:
-            last_sent_stream_id = await send(
-                etype=EventTypes.JoinRules, content={"join_rule": config["join_rules"]}
+            join_rules_event, join_rules_context = await create_event(
+                EventTypes.JoinRules,
+                {"join_rule": config["join_rules"]},
+                True,
             )
+            current_state_group = join_rules_context._state_group
+            events_to_send.append((join_rules_event, join_rules_context))
 
         if (EventTypes.RoomHistoryVisibility, "") not in initial_state:
-            last_sent_stream_id = await send(
-                etype=EventTypes.RoomHistoryVisibility,
-                content={"history_visibility": config["history_visibility"]},
+            visibility_event, visibility_context = await create_event(
+                EventTypes.RoomHistoryVisibility,
+                {"history_visibility": config["history_visibility"]},
+                True,
             )
+            current_state_group = visibility_context._state_group
+            events_to_send.append((visibility_event, visibility_context))
 
         if config["guest_can_join"]:
             if (EventTypes.GuestAccess, "") not in initial_state:
-                last_sent_stream_id = await send(
-                    etype=EventTypes.GuestAccess,
-                    content={EventContentFields.GUEST_ACCESS: GuestAccess.CAN_JOIN},
+                guest_access_event, guest_access_context = await create_event(
+                    EventTypes.GuestAccess,
+                    {EventContentFields.GUEST_ACCESS: GuestAccess.CAN_JOIN},
+                    True,
                 )
+                current_state_group = guest_access_context._state_group
+                events_to_send.append((guest_access_event, guest_access_context))
 
         for (etype, state_key), content in initial_state.items():
-            last_sent_stream_id = await send(
-                etype=etype, state_key=state_key, content=content
+            event, context = await create_event(
+                etype, content, True, state_key=state_key
             )
+            current_state_group = context._state_group
+            events_to_send.append((event, context))
 
         if config["encrypted"]:
-            last_sent_stream_id = await send(
-                etype=EventTypes.RoomEncryption,
+            encryption_event, encryption_context = await create_event(
+                EventTypes.RoomEncryption,
+                {"algorithm": RoomEncryptionAlgorithms.DEFAULT},
+                True,
                 state_key="",
-                content={"algorithm": RoomEncryptionAlgorithms.DEFAULT},
             )
+            events_to_send.append((encryption_event, encryption_context))
 
-        return last_sent_stream_id, last_sent_event_id, depth
+        last_event = await self.event_creation_handler.handle_new_client_event(
+            creator, events_to_send, ignore_shadow_ban=True
+        )
+        assert last_event.internal_metadata.stream_ordering is not None
+        return last_event.internal_metadata.stream_ordering, last_event.event_id, depth
 
     def _generate_room_id(self) -> str:
         """Generates a random room ID.
@@ -1476,7 +1540,9 @@ class TimestampLookupHandler:
             )
 
             likely_domains = (
-                await self._storage_controllers.state.get_current_hosts_in_room(room_id)
+                await self._storage_controllers.state.get_current_hosts_in_room_ordered(
+                    room_id
+                )
             )
 
             # Loop through each homeserver candidate until we get a succesful response
diff --git a/synapse/handlers/room_batch.py b/synapse/handlers/room_batch.py
index 1414e575d6..411a6fb22f 100644
--- a/synapse/handlers/room_batch.py
+++ b/synapse/handlers/room_batch.py
@@ -379,8 +379,7 @@ class RoomBatchHandler:
                 await self.create_requester_for_user_id_from_app_service(
                     event.sender, app_service_requester.app_service
                 ),
-                event=event,
-                context=context,
+                events_and_context=[(event, context)],
             )
 
         return event_ids
diff --git a/synapse/handlers/room_member.py b/synapse/handlers/room_member.py
index e0d0a8941c..fbf874da15 100644
--- a/synapse/handlers/room_member.py
+++ b/synapse/handlers/room_member.py
@@ -432,8 +432,7 @@ class RoomMemberHandler(metaclass=abc.ABCMeta):
         with tracing.start_active_span("handle_new_client_event"):
             result_event = await self.event_creation_handler.handle_new_client_event(
                 requester,
-                event,
-                context,
+                events_and_context=[(event, context)],
                 extra_users=[target],
                 ratelimit=ratelimit,
             )
@@ -1252,7 +1251,10 @@ class RoomMemberHandler(metaclass=abc.ABCMeta):
                 raise SynapseError(403, "This room has been blocked on this server")
 
         event = await self.event_creation_handler.handle_new_client_event(
-            requester, event, context, extra_users=[target_user], ratelimit=ratelimit
+            requester,
+            events_and_context=[(event, context)],
+            extra_users=[target_user],
+            ratelimit=ratelimit,
         )
 
         prev_member_event_id = prev_state_ids.get(
@@ -1860,8 +1862,7 @@ class RoomMemberMasterHandler(RoomMemberHandler):
 
         result_event = await self.event_creation_handler.handle_new_client_event(
             requester,
-            event,
-            context,
+            events_and_context=[(event, context)],
             extra_users=[UserID.from_string(target_user)],
         )
         # we know it was persisted, so must have a stream ordering
diff --git a/synapse/handlers/send_email.py b/synapse/handlers/send_email.py
index e2844799e8..804cc6e81e 100644
--- a/synapse/handlers/send_email.py
+++ b/synapse/handlers/send_email.py
@@ -187,6 +187,19 @@ class SendEmailHandler:
         multipart_msg["To"] = email_address
         multipart_msg["Date"] = email.utils.formatdate()
         multipart_msg["Message-ID"] = email.utils.make_msgid()
+        # Discourage automatic responses to Synapse's emails.
+        # Per RFC 3834, automatic responses should not be sent if the "Auto-Submitted"
+        # header is present with any value other than "no". See
+        #     https://www.rfc-editor.org/rfc/rfc3834.html#section-5.1
+        multipart_msg["Auto-Submitted"] = "auto-generated"
+        # Also include a Microsoft-Exchange specific header:
+        #    https://learn.microsoft.com/en-us/openspecs/exchange_server_protocols/ms-oxcmail/ced68690-498a-4567-9d14-5c01f974d8b1
+        # which suggests it can take the value "All" to "suppress all auto-replies",
+        # or a comma separated list of auto-reply classes to suppress.
+        # The following stack overflow question has a little more context:
+        #    https://stackoverflow.com/a/25324691/5252017
+        #    https://stackoverflow.com/a/61646381/5252017
+        multipart_msg["X-Auto-Response-Suppress"] = "All"
         multipart_msg.attach(text_part)
         multipart_msg.attach(html_part)
 
diff --git a/synapse/handlers/sso.py b/synapse/handlers/sso.py
index 6bc1cbd787..e035677b8a 100644
--- a/synapse/handlers/sso.py
+++ b/synapse/handlers/sso.py
@@ -147,6 +147,9 @@ class UsernameMappingSession:
     # A unique identifier for this SSO provider, e.g.  "oidc" or "saml".
     auth_provider_id: str
 
+    # An optional session ID from the IdP.
+    auth_provider_session_id: Optional[str]
+
     # user ID on the IdP server
     remote_user_id: str
 
@@ -464,6 +467,7 @@ class SsoHandler:
                         client_redirect_url,
                         next_step_url,
                         extra_login_attributes,
+                        auth_provider_session_id,
                     )
 
                 user_id = await self._register_mapped_user(
@@ -585,6 +589,7 @@ class SsoHandler:
         client_redirect_url: str,
         next_step_url: bytes,
         extra_login_attributes: Optional[JsonDict],
+        auth_provider_session_id: Optional[str],
     ) -> NoReturn:
         """Creates a UsernameMappingSession and redirects the browser
 
@@ -607,6 +612,8 @@ class SsoHandler:
             extra_login_attributes: An optional dictionary of extra
                 attributes to be provided to the client in the login response.
 
+            auth_provider_session_id: An optional session ID from the IdP.
+
         Raises:
             RedirectException
         """
@@ -615,6 +622,7 @@ class SsoHandler:
         now = self._clock.time_msec()
         session = UsernameMappingSession(
             auth_provider_id=auth_provider_id,
+            auth_provider_session_id=auth_provider_session_id,
             remote_user_id=remote_user_id,
             display_name=attributes.display_name,
             emails=attributes.emails,
@@ -968,6 +976,7 @@ class SsoHandler:
             session.client_redirect_url,
             session.extra_login_attributes,
             new_user=True,
+            auth_provider_session_id=session.auth_provider_session_id,
         )
 
     def _expire_old_sessions(self) -> None:
diff --git a/synapse/handlers/sync.py b/synapse/handlers/sync.py
index f7fd6d7933..fed9701ece 100644
--- a/synapse/handlers/sync.py
+++ b/synapse/handlers/sync.py
@@ -1196,7 +1196,9 @@ class SyncHandler:
             room_id: The partial state room to find the remaining memberships for.
             members_to_fetch: The memberships to find.
             events_with_membership_auth: A mapping from user IDs to events whose auth
-                events are known to contain their membership.
+                events would contain their prior membership, if one exists.
+                Note that join events will not cite a prior membership if a user has
+                never been in a room before.
             found_state_ids: A dict from (type, state_key) -> state_event_id, containing
                 memberships that have been previously found. Entries in
                 `members_to_fetch` that have a membership in `found_state_ids` are
@@ -1206,6 +1208,10 @@ class SyncHandler:
             A dict from ("m.room.member", state_key) -> state_event_id, containing the
             memberships missing from `found_state_ids`.
 
+            When `events_with_membership_auth` contains a join event for a given user
+            which does not cite a prior membership, no membership is returned for that
+            user.
+
         Raises:
             KeyError: if `events_with_membership_auth` does not have an entry for a
                 missing membership. Memberships in `found_state_ids` do not need an
@@ -1223,8 +1229,18 @@ class SyncHandler:
             if (EventTypes.Member, member) in found_state_ids:
                 continue
 
-            missing_members.add(member)
             event_with_membership_auth = events_with_membership_auth[member]
+            is_join = (
+                event_with_membership_auth.is_state()
+                and event_with_membership_auth.type == EventTypes.Member
+                and event_with_membership_auth.state_key == member
+                and event_with_membership_auth.content.get("membership")
+                == Membership.JOIN
+            )
+            if not is_join:
+                # The event must include the desired membership as an auth event, unless
+                # it's the first join event for a given user.
+                missing_members.add(member)
             auth_event_ids.update(event_with_membership_auth.auth_event_ids())
 
         auth_events = await self.store.get_events(auth_event_ids)
@@ -1248,7 +1264,7 @@ class SyncHandler:
                     auth_event.type == EventTypes.Member
                     and auth_event.state_key == member
                 ):
-                    missing_members.remove(member)
+                    missing_members.discard(member)
                     additional_state_ids[
                         (EventTypes.Member, member)
                     ] = auth_event.event_id
@@ -1479,16 +1495,14 @@ class SyncHandler:
                 since_token.device_list_key
             )
             if changed_users is not None:
-                result = await self.store.get_rooms_for_users_with_stream_ordering(
-                    changed_users
-                )
+                result = await self.store.get_rooms_for_users(changed_users)
 
                 for changed_user_id, entries in result.items():
                     # Check if the changed user shares any rooms with the user,
                     # or if the changed user is the syncing user (as we always
                     # want to include device list updates of their own devices).
                     if user_id == changed_user_id or any(
-                        e.room_id in joined_rooms for e in entries
+                        rid in joined_rooms for rid in entries
                     ):
                         users_that_have_changed.add(changed_user_id)
             else:
@@ -1522,13 +1536,9 @@ class SyncHandler:
                 newly_left_users.update(left_users)
 
             # Remove any users that we still share a room with.
-            left_users_rooms = (
-                await self.store.get_rooms_for_users_with_stream_ordering(
-                    newly_left_users
-                )
-            )
+            left_users_rooms = await self.store.get_rooms_for_users(newly_left_users)
             for user_id, entries in left_users_rooms.items():
-                if any(e.room_id in joined_rooms for e in entries):
+                if any(rid in joined_rooms for rid in entries):
                     newly_left_users.discard(user_id)
 
             return DeviceListUpdates(
diff --git a/synapse/handlers/user_directory.py b/synapse/handlers/user_directory.py
index 8c3c52e1ca..3610b6bf78 100644
--- a/synapse/handlers/user_directory.py
+++ b/synapse/handlers/user_directory.py
@@ -13,7 +13,7 @@
 # limitations under the License.
 
 import logging
-from typing import TYPE_CHECKING, Any, Dict, List, Optional
+from typing import TYPE_CHECKING, Any, Dict, List, Optional, Set, Tuple
 
 import synapse.metrics
 from synapse.api.constants import EventTypes, HistoryVisibility, JoinRules, Membership
@@ -379,7 +379,7 @@ class UserDirectoryHandler(StateDeltasHandler):
             user_id, event.content.get("displayname"), event.content.get("avatar_url")
         )
 
-    async def _track_user_joined_room(self, room_id: str, user_id: str) -> None:
+    async def _track_user_joined_room(self, room_id: str, joining_user_id: str) -> None:
         """Someone's just joined a room. Update `users_in_public_rooms` or
         `users_who_share_private_rooms` as appropriate.
 
@@ -390,32 +390,44 @@ class UserDirectoryHandler(StateDeltasHandler):
             room_id
         )
         if is_public:
-            await self.store.add_users_in_public_rooms(room_id, (user_id,))
+            await self.store.add_users_in_public_rooms(room_id, (joining_user_id,))
         else:
             users_in_room = await self.store.get_users_in_room(room_id)
             other_users_in_room = [
                 other
                 for other in users_in_room
-                if other != user_id
+                if other != joining_user_id
                 and (
+                    # We can't apply any special rules to remote users so
+                    # they're always included
                     not self.is_mine_id(other)
+                    # Check the special rules whether the local user should be
+                    # included in the user directory
                     or await self.store.should_include_local_user_in_dir(other)
                 )
             ]
-            to_insert = set()
+            updates_to_users_who_share_rooms: Set[Tuple[str, str]] = set()
 
-            # First, if they're our user then we need to update for every user
-            if self.is_mine_id(user_id):
+            # First, if the joining user is our local user then we need an
+            # update for every other user in the room.
+            if self.is_mine_id(joining_user_id):
                 for other_user_id in other_users_in_room:
-                    to_insert.add((user_id, other_user_id))
+                    updates_to_users_who_share_rooms.add(
+                        (joining_user_id, other_user_id)
+                    )
 
-            # Next we need to update for every local user in the room
+            # Next, we need an update for every other local user in the room
+            # that they now share a room with the joining user.
             for other_user_id in other_users_in_room:
                 if self.is_mine_id(other_user_id):
-                    to_insert.add((other_user_id, user_id))
+                    updates_to_users_who_share_rooms.add(
+                        (other_user_id, joining_user_id)
+                    )
 
-            if to_insert:
-                await self.store.add_users_who_share_private_room(room_id, to_insert)
+            if updates_to_users_who_share_rooms:
+                await self.store.add_users_who_share_private_room(
+                    room_id, updates_to_users_who_share_rooms
+                )
 
     async def _handle_remove_user(self, room_id: str, user_id: str) -> None:
         """Called when when someone leaves a room. The user may be local or remote.
diff --git a/synapse/logging/context.py b/synapse/logging/context.py
index a417b13ffd..9bbe07f21d 100644
--- a/synapse/logging/context.py
+++ b/synapse/logging/context.py
@@ -579,7 +579,7 @@ class LoggingContextFilter(logging.Filter):
             True to include the record in the log output.
         """
         context = current_context()
-        record.request = self._default_request  # type: ignore
+        record.request = self._default_request
 
         # context should never be None, but if it somehow ends up being, then
         # we end up in a death spiral of infinite loops, so let's check, for
@@ -587,21 +587,21 @@ class LoggingContextFilter(logging.Filter):
         if context is not None:
             # Logging is interested in the request ID. Note that for backwards
             # compatibility this is stored as the "request" on the record.
-            record.request = str(context)  # type: ignore
+            record.request = str(context)
 
             # Add some data from the HTTP request.
             request = context.request
             if request is None:
                 return True
 
-            record.ip_address = request.ip_address  # type: ignore
-            record.site_tag = request.site_tag  # type: ignore
-            record.requester = request.requester  # type: ignore
-            record.authenticated_entity = request.authenticated_entity  # type: ignore
-            record.method = request.method  # type: ignore
-            record.url = request.url  # type: ignore
-            record.protocol = request.protocol  # type: ignore
-            record.user_agent = request.user_agent  # type: ignore
+            record.ip_address = request.ip_address
+            record.site_tag = request.site_tag
+            record.requester = request.requester
+            record.authenticated_entity = request.authenticated_entity
+            record.method = request.method
+            record.url = request.url
+            record.protocol = request.protocol
+            record.user_agent = request.user_agent
 
         return True
 
diff --git a/synapse/logging/tracing.py b/synapse/logging/tracing.py
index d594ab735b..d455854467 100644
--- a/synapse/logging/tracing.py
+++ b/synapse/logging/tracing.py
@@ -209,7 +209,10 @@ class _DummyLookup(object):
     def __init__(self, value: T) -> None:
         self.value = value
 
-    def __getattribute__(self, name: str) -> T:
+    # type-ignore: Because mypy says "A function returning TypeVar should receive at
+    # least one argument containing the same Typevar" but this is just a dummy
+    # stand-in that doesn't need any input.
+    def __getattribute__(self, name: str) -> T:  # type: ignore[type-var]
         return object.__getattribute__(self, "value")
 
 
@@ -949,9 +952,9 @@ def tag_args(func: Callable[P, R]) -> Callable[P, R]:
         # FIXME: We could update this to handle any type of function by ignoring the
         #   first argument only if it's named `self` or `cls`. This isn't fool-proof
         #   but handles the idiomatic cases.
-        for i, arg in enumerate(args[1:], start=1):  # type: ignore[index]
+        for i, arg in enumerate(args[1:], start=1):
             set_attribute(SynapseTags.FUNC_ARG_PREFIX + argspec.args[i], str(arg))
-        set_attribute(SynapseTags.FUNC_ARGS, str(args[len(argspec.args) :]))  # type: ignore[index]
+        set_attribute(SynapseTags.FUNC_ARGS, str(args[len(argspec.args) :]))
         set_attribute(SynapseTags.FUNC_KWARGS, str(kwargs))
         yield
 
diff --git a/synapse/module_api/__init__.py b/synapse/module_api/__init__.py
index 59755bff6d..b7b2d3b8c5 100644
--- a/synapse/module_api/__init__.py
+++ b/synapse/module_api/__init__.py
@@ -842,6 +842,8 @@ class ModuleApi:
         however invalidation that needs to go to other workers needs to call `invalidate_cache`
         on the module API instead.
 
+        Added in Synapse v1.69.0.
+
         Args:
             cached_function: The cached function that will be registered to receive invalidation
             locally and from other workers.
@@ -856,6 +858,8 @@ class ModuleApi:
         """Invalidate a cache entry of a cached function across workers. The cached function
         needs to be registered on all workers first with `register_cached_function`.
 
+        Added in Synapse v1.69.0.
+
         Args:
             cached_function: The cached function that needs an invalidation
             keys: keys of the entry to invalidate, usually matching the arguments of the
diff --git a/synapse/push/bulk_push_rule_evaluator.py b/synapse/push/bulk_push_rule_evaluator.py
index 404379ef67..4270438918 100644
--- a/synapse/push/bulk_push_rule_evaluator.py
+++ b/synapse/push/bulk_push_rule_evaluator.py
@@ -17,6 +17,7 @@ import itertools
 import logging
 from typing import (
     TYPE_CHECKING,
+    Any,
     Collection,
     Dict,
     Iterable,
@@ -37,13 +38,11 @@ from synapse.events.snapshot import EventContext
 from synapse.state import POWER_KEY
 from synapse.storage.databases.main.roommember import EventIdMembership
 from synapse.storage.state import StateFilter
-from synapse.synapse_rust.push import FilteredPushRules, PushRule
+from synapse.synapse_rust.push import FilteredPushRules, PushRule, PushRuleEvaluator
 from synapse.util.caches import register_cache
 from synapse.util.metrics import measure_func
 from synapse.visibility import filter_event_for_clients_with_state
 
-from .push_rule_evaluator import PushRuleEvaluatorForEvent
-
 if TYPE_CHECKING:
     from synapse.server import HomeServer
 
@@ -173,7 +172,11 @@ class BulkPushRuleEvaluator:
 
     async def _get_power_levels_and_sender_level(
         self, event: EventBase, context: EventContext
-    ) -> Tuple[dict, int]:
+    ) -> Tuple[dict, Optional[int]]:
+        # There are no power levels and sender levels possible to get from outlier
+        if event.internal_metadata.is_outlier():
+            return {}, None
+
         event_types = auth_types_for_event(event.room_version, event)
         prev_state_ids = await context.get_prev_state_ids(
             StateFilter.from_types(event_types)
@@ -250,8 +253,8 @@ class BulkPushRuleEvaluator:
         should increment the unread count, and insert the results into the
         event_push_actions_staging table.
         """
-        if event.internal_metadata.is_outlier():
-            # This can happen due to out of band memberships
+        if not event.internal_metadata.is_notifiable():
+            # Push rules for events that aren't notifiable can't be processed by this
             return
 
         # Disable counting as unread unless the experimental configuration is
@@ -286,11 +289,11 @@ class BulkPushRuleEvaluator:
             if relation.rel_type == RelationTypes.THREAD:
                 thread_id = relation.parent_id
 
-        evaluator = PushRuleEvaluatorForEvent(
-            event,
+        evaluator = PushRuleEvaluator(
+            _flatten_dict(event),
             room_member_count,
             sender_power_level,
-            power_levels,
+            power_levels.get("notifications", {}),
             relations,
             self._relations_match_enabled,
         )
@@ -300,20 +303,10 @@ class BulkPushRuleEvaluator:
             event.room_id, users
         )
 
-        # This is a check for the case where user joins a room without being
-        # allowed to see history, and then the server receives a delayed event
-        # from before the user joined, which they should not be pushed for
-        uids_with_visibility = await filter_event_for_clients_with_state(
-            self.store, users, event, context
-        )
-
         for uid, rules in rules_by_user.items():
             if event.sender == uid:
                 continue
 
-            if uid not in uids_with_visibility:
-                continue
-
             display_name = None
             profile = profiles.get(uid)
             if profile:
@@ -334,17 +327,30 @@ class BulkPushRuleEvaluator:
                 # current user, it'll be added to the dict later.
                 actions_by_user[uid] = []
 
-            for rule, enabled in rules.rules():
-                if not enabled:
-                    continue
+            actions = evaluator.run(rules, uid, display_name)
+            if "notify" in actions:
+                # Push rules say we should notify the user of this event
+                actions_by_user[uid] = actions
+
+        # If there aren't any actions then we can skip the rest of the
+        # processing.
+        if not actions_by_user:
+            return
+
+        # This is a check for the case where user joins a room without being
+        # allowed to see history, and then the server receives a delayed event
+        # from before the user joined, which they should not be pushed for
+        #
+        # We do this *after* calculating the push actions as a) its unlikely
+        # that we'll filter anyone out and b) for large rooms its likely that
+        # most users will have push disabled and so the set of users to check is
+        # much smaller.
+        uids_with_visibility = await filter_event_for_clients_with_state(
+            self.store, actions_by_user.keys(), event, context
+        )
 
-                matches = evaluator.check_conditions(rule.conditions, uid, display_name)
-                if matches:
-                    actions = [x for x in rule.actions if x != "dont_notify"]
-                    if actions and "notify" in actions:
-                        # Push rules say we should notify the user of this event
-                        actions_by_user[uid] = actions
-                    break
+        for user_id in set(actions_by_user).difference(uids_with_visibility):
+            actions_by_user.pop(user_id, None)
 
         # Mark in the DB staging area the push actions for users who should be
         # notified for this event. (This will then get handled when we persist
@@ -361,3 +367,21 @@ MemberMap = Dict[str, Optional[EventIdMembership]]
 Rule = Dict[str, dict]
 RulesByUser = Dict[str, List[Rule]]
 StateGroup = Union[object, int]
+
+
+def _flatten_dict(
+    d: Union[EventBase, Mapping[str, Any]],
+    prefix: Optional[List[str]] = None,
+    result: Optional[Dict[str, str]] = None,
+) -> Dict[str, str]:
+    if prefix is None:
+        prefix = []
+    if result is None:
+        result = {}
+    for key, value in d.items():
+        if isinstance(value, str):
+            result[".".join(prefix + [key])] = value.lower()
+        elif isinstance(value, Mapping):
+            _flatten_dict(value, prefix=(prefix + [key]), result=result)
+
+    return result
diff --git a/synapse/push/clientformat.py b/synapse/push/clientformat.py
index ebc13beda1..7095ae83f9 100644
--- a/synapse/push/clientformat.py
+++ b/synapse/push/clientformat.py
@@ -102,10 +102,8 @@ def _rule_to_template(rule: PushRule) -> Optional[Dict[str, Any]]:
         # with PRIORITY_CLASS_INVERSE_MAP.
         raise ValueError("Unexpected template_name: %s" % (template_name,))
 
-    if unscoped_rule_id:
-        templaterule["rule_id"] = unscoped_rule_id
-    if rule.default:
-        templaterule["default"] = True
+    templaterule["rule_id"] = unscoped_rule_id
+    templaterule["default"] = rule.default
     return templaterule
 
 
diff --git a/synapse/push/httppusher.py b/synapse/push/httppusher.py
index 11299367d2..3149492ede 100644
--- a/synapse/push/httppusher.py
+++ b/synapse/push/httppusher.py
@@ -14,7 +14,7 @@
 # limitations under the License.
 import logging
 import urllib.parse
-from typing import TYPE_CHECKING, Any, Dict, Iterable, Optional, Union
+from typing import TYPE_CHECKING, Any, Dict, Iterable, List, Optional, Union
 
 from prometheus_client import Counter
 
@@ -28,7 +28,7 @@ from synapse.metrics.background_process_metrics import run_as_background_process
 from synapse.push import Pusher, PusherConfig, PusherConfigException
 from synapse.storage.databases.main.event_push_actions import HttpPushAction
 
-from . import push_rule_evaluator, push_tools
+from . import push_tools
 
 if TYPE_CHECKING:
     from synapse.server import HomeServer
@@ -56,6 +56,39 @@ http_badges_failed_counter = Counter(
 )
 
 
+def tweaks_for_actions(actions: List[Union[str, Dict]]) -> Dict[str, Any]:
+    """
+    Converts a list of actions into a `tweaks` dict (which can then be passed to
+        the push gateway).
+
+    This function ignores all actions other than `set_tweak` actions, and treats
+    absent `value`s as `True`, which agrees with the only spec-defined treatment
+    of absent `value`s (namely, for `highlight` tweaks).
+
+    Args:
+        actions: list of actions
+            e.g. [
+                {"set_tweak": "a", "value": "AAA"},
+                {"set_tweak": "b", "value": "BBB"},
+                {"set_tweak": "highlight"},
+                "notify"
+            ]
+
+    Returns:
+        dictionary of tweaks for those actions
+            e.g. {"a": "AAA", "b": "BBB", "highlight": True}
+    """
+    tweaks = {}
+    for a in actions:
+        if not isinstance(a, dict):
+            continue
+        if "set_tweak" in a:
+            # value is allowed to be absent in which case the value assumed
+            # should be True.
+            tweaks[a["set_tweak"]] = a.get("value", True)
+    return tweaks
+
+
 class HttpPusher(Pusher):
     INITIAL_BACKOFF_SEC = 1  # in seconds because that's what Twisted takes
     MAX_BACKOFF_SEC = 60 * 60
@@ -281,7 +314,7 @@ class HttpPusher(Pusher):
         if "notify" not in push_action.actions:
             return True
 
-        tweaks = push_rule_evaluator.tweaks_for_actions(push_action.actions)
+        tweaks = tweaks_for_actions(push_action.actions)
         badge = await push_tools.get_badge_count(
             self.hs.get_datastores().main,
             self.user_id,
diff --git a/synapse/push/push_rule_evaluator.py b/synapse/push/push_rule_evaluator.py
deleted file mode 100644
index 3c5632cd91..0000000000
--- a/synapse/push/push_rule_evaluator.py
+++ /dev/null
@@ -1,361 +0,0 @@
-# Copyright 2015, 2016 OpenMarket Ltd
-# Copyright 2017 New Vector Ltd
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import logging
-import re
-from typing import (
-    Any,
-    Dict,
-    List,
-    Mapping,
-    Optional,
-    Pattern,
-    Sequence,
-    Set,
-    Tuple,
-    Union,
-)
-
-from matrix_common.regex import glob_to_regex, to_word_pattern
-
-from synapse.events import EventBase
-from synapse.types import UserID
-from synapse.util.caches.lrucache import LruCache
-
-logger = logging.getLogger(__name__)
-
-
-GLOB_REGEX = re.compile(r"\\\[(\\\!|)(.*)\\\]")
-IS_GLOB = re.compile(r"[\?\*\[\]]")
-INEQUALITY_EXPR = re.compile("^([=<>]*)([0-9]*)$")
-
-
-def _room_member_count(
-    ev: EventBase, condition: Mapping[str, Any], room_member_count: int
-) -> bool:
-    return _test_ineq_condition(condition, room_member_count)
-
-
-def _sender_notification_permission(
-    ev: EventBase,
-    condition: Mapping[str, Any],
-    sender_power_level: int,
-    power_levels: Dict[str, Union[int, Dict[str, int]]],
-) -> bool:
-    notif_level_key = condition.get("key")
-    if notif_level_key is None:
-        return False
-
-    notif_levels = power_levels.get("notifications", {})
-    assert isinstance(notif_levels, dict)
-    room_notif_level = notif_levels.get(notif_level_key, 50)
-
-    return sender_power_level >= room_notif_level
-
-
-def _test_ineq_condition(condition: Mapping[str, Any], number: int) -> bool:
-    if "is" not in condition:
-        return False
-    m = INEQUALITY_EXPR.match(condition["is"])
-    if not m:
-        return False
-    ineq = m.group(1)
-    rhs = m.group(2)
-    if not rhs.isdigit():
-        return False
-    rhs_int = int(rhs)
-
-    if ineq == "" or ineq == "==":
-        return number == rhs_int
-    elif ineq == "<":
-        return number < rhs_int
-    elif ineq == ">":
-        return number > rhs_int
-    elif ineq == ">=":
-        return number >= rhs_int
-    elif ineq == "<=":
-        return number <= rhs_int
-    else:
-        return False
-
-
-def tweaks_for_actions(actions: List[Union[str, Dict]]) -> Dict[str, Any]:
-    """
-    Converts a list of actions into a `tweaks` dict (which can then be passed to
-        the push gateway).
-
-    This function ignores all actions other than `set_tweak` actions, and treats
-    absent `value`s as `True`, which agrees with the only spec-defined treatment
-    of absent `value`s (namely, for `highlight` tweaks).
-
-    Args:
-        actions: list of actions
-            e.g. [
-                {"set_tweak": "a", "value": "AAA"},
-                {"set_tweak": "b", "value": "BBB"},
-                {"set_tweak": "highlight"},
-                "notify"
-            ]
-
-    Returns:
-        dictionary of tweaks for those actions
-            e.g. {"a": "AAA", "b": "BBB", "highlight": True}
-    """
-    tweaks = {}
-    for a in actions:
-        if not isinstance(a, dict):
-            continue
-        if "set_tweak" in a:
-            # value is allowed to be absent in which case the value assumed
-            # should be True.
-            tweaks[a["set_tweak"]] = a.get("value", True)
-    return tweaks
-
-
-class PushRuleEvaluatorForEvent:
-    def __init__(
-        self,
-        event: EventBase,
-        room_member_count: int,
-        sender_power_level: int,
-        power_levels: Dict[str, Union[int, Dict[str, int]]],
-        relations: Dict[str, Set[Tuple[str, str]]],
-        relations_match_enabled: bool,
-    ):
-        self._event = event
-        self._room_member_count = room_member_count
-        self._sender_power_level = sender_power_level
-        self._power_levels = power_levels
-        self._relations = relations
-        self._relations_match_enabled = relations_match_enabled
-
-        # Maps strings of e.g. 'content.body' -> event["content"]["body"]
-        self._value_cache = _flatten_dict(event)
-
-        # Maps cache keys to final values.
-        self._condition_cache: Dict[str, bool] = {}
-
-    def check_conditions(
-        self, conditions: Sequence[Mapping], uid: str, display_name: Optional[str]
-    ) -> bool:
-        """
-        Returns true if a user's conditions/user ID/display name match the event.
-
-        Args:
-            conditions: The user's conditions to match.
-            uid: The user's MXID.
-            display_name: The display name.
-
-        Returns:
-             True if all conditions match the event, False otherwise.
-        """
-        for cond in conditions:
-            _cache_key = cond.get("_cache_key", None)
-            if _cache_key:
-                res = self._condition_cache.get(_cache_key, None)
-                if res is False:
-                    return False
-                elif res is True:
-                    continue
-
-            res = self.matches(cond, uid, display_name)
-            if _cache_key:
-                self._condition_cache[_cache_key] = bool(res)
-
-            if not res:
-                return False
-
-        return True
-
-    def matches(
-        self, condition: Mapping[str, Any], user_id: str, display_name: Optional[str]
-    ) -> bool:
-        """
-        Returns true if a user's condition/user ID/display name match the event.
-
-        Args:
-            condition: The user's condition to match.
-            uid: The user's MXID.
-            display_name: The display name, or None if there is not one.
-
-        Returns:
-             True if the condition matches the event, False otherwise.
-        """
-        if condition["kind"] == "event_match":
-            return self._event_match(condition, user_id)
-        elif condition["kind"] == "contains_display_name":
-            return self._contains_display_name(display_name)
-        elif condition["kind"] == "room_member_count":
-            return _room_member_count(self._event, condition, self._room_member_count)
-        elif condition["kind"] == "sender_notification_permission":
-            return _sender_notification_permission(
-                self._event, condition, self._sender_power_level, self._power_levels
-            )
-        elif (
-            condition["kind"] == "org.matrix.msc3772.relation_match"
-            and self._relations_match_enabled
-        ):
-            return self._relation_match(condition, user_id)
-        else:
-            # XXX This looks incorrect -- we have reached an unknown condition
-            #     kind and are unconditionally returning that it matches. Note
-            #     that it seems possible to provide a condition to the /pushrules
-            #     endpoint with an unknown kind, see _rule_tuple_from_request_object.
-            return True
-
-    def _event_match(self, condition: Mapping, user_id: str) -> bool:
-        """
-        Check an "event_match" push rule condition.
-
-        Args:
-            condition: The "event_match" push rule condition to match.
-            user_id: The user's MXID.
-
-        Returns:
-             True if the condition matches the event, False otherwise.
-        """
-        pattern = condition.get("pattern", None)
-
-        if not pattern:
-            pattern_type = condition.get("pattern_type", None)
-            if pattern_type == "user_id":
-                pattern = user_id
-            elif pattern_type == "user_localpart":
-                pattern = UserID.from_string(user_id).localpart
-
-        if not pattern:
-            logger.warning("event_match condition with no pattern")
-            return False
-
-        # XXX: optimisation: cache our pattern regexps
-        if condition["key"] == "content.body":
-            body = self._event.content.get("body", None)
-            if not body or not isinstance(body, str):
-                return False
-
-            return _glob_matches(pattern, body, word_boundary=True)
-        else:
-            haystack = self._value_cache.get(condition["key"], None)
-            if haystack is None:
-                return False
-
-            return _glob_matches(pattern, haystack)
-
-    def _contains_display_name(self, display_name: Optional[str]) -> bool:
-        """
-        Check an "event_match" push rule condition.
-
-        Args:
-            display_name: The display name, or None if there is not one.
-
-        Returns:
-             True if the display name is found in the event body, False otherwise.
-        """
-        if not display_name:
-            return False
-
-        body = self._event.content.get("body", None)
-        if not body or not isinstance(body, str):
-            return False
-
-        # Similar to _glob_matches, but do not treat display_name as a glob.
-        r = regex_cache.get((display_name, False, True), None)
-        if not r:
-            r1 = re.escape(display_name)
-            r1 = to_word_pattern(r1)
-            r = re.compile(r1, flags=re.IGNORECASE)
-            regex_cache[(display_name, False, True)] = r
-
-        return bool(r.search(body))
-
-    def _relation_match(self, condition: Mapping, user_id: str) -> bool:
-        """
-        Check an "relation_match" push rule condition.
-
-        Args:
-            condition: The "event_match" push rule condition to match.
-            user_id: The user's MXID.
-
-        Returns:
-             True if the condition matches the event, False otherwise.
-        """
-        rel_type = condition.get("rel_type")
-        if not rel_type:
-            logger.warning("relation_match condition missing rel_type")
-            return False
-
-        sender_pattern = condition.get("sender")
-        if sender_pattern is None:
-            sender_type = condition.get("sender_type")
-            if sender_type == "user_id":
-                sender_pattern = user_id
-        type_pattern = condition.get("type")
-
-        # If any other relations matches, return True.
-        for sender, event_type in self._relations.get(rel_type, ()):
-            if sender_pattern and not _glob_matches(sender_pattern, sender):
-                continue
-            if type_pattern and not _glob_matches(type_pattern, event_type):
-                continue
-            # All values must have matched.
-            return True
-
-        # No relations matched.
-        return False
-
-
-# Caches (string, is_glob, word_boundary) -> regex for push. See _glob_matches
-regex_cache: LruCache[Tuple[str, bool, bool], Pattern] = LruCache(
-    50000, "regex_push_cache"
-)
-
-
-def _glob_matches(glob: str, value: str, word_boundary: bool = False) -> bool:
-    """Tests if value matches glob.
-
-    Args:
-        glob
-        value: String to test against glob.
-        word_boundary: Whether to match against word boundaries or entire
-            string. Defaults to False.
-    """
-
-    try:
-        r = regex_cache.get((glob, True, word_boundary), None)
-        if not r:
-            r = glob_to_regex(glob, word_boundary=word_boundary)
-            regex_cache[(glob, True, word_boundary)] = r
-        return bool(r.search(value))
-    except re.error:
-        logger.warning("Failed to parse glob to regex: %r", glob)
-        return False
-
-
-def _flatten_dict(
-    d: Union[EventBase, Mapping[str, Any]],
-    prefix: Optional[List[str]] = None,
-    result: Optional[Dict[str, str]] = None,
-) -> Dict[str, str]:
-    if prefix is None:
-        prefix = []
-    if result is None:
-        result = {}
-    for key, value in d.items():
-        if isinstance(value, str):
-            result[".".join(prefix + [key])] = value.lower()
-        elif isinstance(value, Mapping):
-            _flatten_dict(value, prefix=(prefix + [key]), result=result)
-
-    return result
diff --git a/synapse/replication/http/__init__.py b/synapse/replication/http/__init__.py
index 53aa7fa4c6..ac9a92240a 100644
--- a/synapse/replication/http/__init__.py
+++ b/synapse/replication/http/__init__.py
@@ -25,6 +25,7 @@ from synapse.replication.http import (
     push,
     register,
     send_event,
+    send_events,
     state,
     streams,
 )
@@ -43,6 +44,7 @@ class ReplicationRestResource(JsonResource):
 
     def register_servlets(self, hs: "HomeServer") -> None:
         send_event.register_servlets(hs, self)
+        send_events.register_servlets(hs, self)
         federation.register_servlets(hs, self)
         presence.register_servlets(hs, self)
         membership.register_servlets(hs, self)
diff --git a/synapse/replication/http/register.py b/synapse/replication/http/register.py
index 6c8f8388fd..61abb529c8 100644
--- a/synapse/replication/http/register.py
+++ b/synapse/replication/http/register.py
@@ -51,6 +51,7 @@ class ReplicationRegisterServlet(ReplicationEndpoint):
         user_type: Optional[str],
         address: Optional[str],
         shadow_banned: bool,
+        approved: bool,
     ) -> JsonDict:
         """
         Args:
@@ -68,6 +69,8 @@ class ReplicationRegisterServlet(ReplicationEndpoint):
                 or None for a normal user.
             address: the IP address used to perform the regitration.
             shadow_banned: Whether to shadow-ban the user
+            approved: Whether the user should be considered already approved by an
+                administrator.
         """
         return {
             "password_hash": password_hash,
@@ -79,6 +82,7 @@ class ReplicationRegisterServlet(ReplicationEndpoint):
             "user_type": user_type,
             "address": address,
             "shadow_banned": shadow_banned,
+            "approved": approved,
         }
 
     async def _handle_request(  # type: ignore[override]
@@ -99,6 +103,7 @@ class ReplicationRegisterServlet(ReplicationEndpoint):
             user_type=content["user_type"],
             address=content["address"],
             shadow_banned=content["shadow_banned"],
+            approved=content["approved"],
         )
 
         return 200, {}
diff --git a/synapse/replication/http/send_event.py b/synapse/replication/http/send_event.py
index 486f04723c..4215a1c1bc 100644
--- a/synapse/replication/http/send_event.py
+++ b/synapse/replication/http/send_event.py
@@ -141,8 +141,8 @@ class ReplicationSendEventRestServlet(ReplicationEndpoint):
             "Got event to send with ID: %s into room: %s", event.event_id, event.room_id
         )
 
-        event = await self.event_creation_handler.persist_and_notify_client_event(
-            requester, event, context, ratelimit=ratelimit, extra_users=extra_users
+        event = await self.event_creation_handler.persist_and_notify_client_events(
+            requester, [(event, context)], ratelimit=ratelimit, extra_users=extra_users
         )
 
         return (
diff --git a/synapse/replication/http/send_events.py b/synapse/replication/http/send_events.py
new file mode 100644
index 0000000000..8889bbb644
--- /dev/null
+++ b/synapse/replication/http/send_events.py
@@ -0,0 +1,171 @@
+# Copyright 2022 The Matrix.org Foundation C.I.C.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import logging
+from typing import TYPE_CHECKING, List, Tuple
+
+from twisted.web.server import Request
+
+from synapse.api.room_versions import KNOWN_ROOM_VERSIONS
+from synapse.events import EventBase, make_event_from_dict
+from synapse.events.snapshot import EventContext
+from synapse.http.server import HttpServer
+from synapse.http.servlet import parse_json_object_from_request
+from synapse.replication.http._base import ReplicationEndpoint
+from synapse.types import JsonDict, Requester, UserID
+from synapse.util.metrics import Measure
+
+if TYPE_CHECKING:
+    from synapse.server import HomeServer
+    from synapse.storage.databases.main import DataStore
+
+logger = logging.getLogger(__name__)
+
+
+class ReplicationSendEventsRestServlet(ReplicationEndpoint):
+    """Handles batches of newly created events on workers, including persisting and
+    notifying.
+
+    The API looks like:
+
+        POST /_synapse/replication/send_events/:txn_id
+
+        {
+            "events": [{
+                "event": { .. serialized event .. },
+                "room_version": .., // "1", "2", "3", etc: the version of the room
+                            // containing the event
+                "event_format_version": .., // 1,2,3 etc: the event format version
+                "internal_metadata": { .. serialized internal_metadata .. },
+                "outlier": true|false,
+                "rejected_reason": ..,   // The event.rejected_reason field
+                "context": { .. serialized event context .. },
+                "requester": { .. serialized requester .. },
+                "ratelimit": true,
+            }]
+        }
+
+        200 OK
+
+        { "stream_id": 12345, "event_id": "$abcdef..." }
+
+    Responds with a 409 when a `PartialStateConflictError` is raised due to an event
+    context that needs to be recomputed due to the un-partial stating of a room.
+
+    """
+
+    NAME = "send_events"
+    PATH_ARGS = ()
+
+    def __init__(self, hs: "HomeServer"):
+        super().__init__(hs)
+
+        self.event_creation_handler = hs.get_event_creation_handler()
+        self.store = hs.get_datastores().main
+        self._storage_controllers = hs.get_storage_controllers()
+        self.clock = hs.get_clock()
+
+    @staticmethod
+    async def _serialize_payload(  # type: ignore[override]
+        events_and_context: List[Tuple[EventBase, EventContext]],
+        store: "DataStore",
+        requester: Requester,
+        ratelimit: bool,
+        extra_users: List[UserID],
+    ) -> JsonDict:
+        """
+        Args:
+            store
+            requester
+            events_and_ctx
+            ratelimit
+        """
+        serialized_events = []
+
+        for event, context in events_and_context:
+            serialized_context = await context.serialize(event, store)
+            serialized_event = {
+                "event": event.get_pdu_json(),
+                "room_version": event.room_version.identifier,
+                "event_format_version": event.format_version,
+                "internal_metadata": event.internal_metadata.get_dict(),
+                "outlier": event.internal_metadata.is_outlier(),
+                "rejected_reason": event.rejected_reason,
+                "context": serialized_context,
+                "requester": requester.serialize(),
+                "ratelimit": ratelimit,
+                "extra_users": [u.to_string() for u in extra_users],
+            }
+            serialized_events.append(serialized_event)
+
+        payload = {"events": serialized_events}
+
+        return payload
+
+    async def _handle_request(  # type: ignore[override]
+        self, request: Request
+    ) -> Tuple[int, JsonDict]:
+        with Measure(self.clock, "repl_send_events_parse"):
+            payload = parse_json_object_from_request(request)
+            events_and_context = []
+            events = payload["events"]
+
+            for event_payload in events:
+                event_dict = event_payload["event"]
+                room_ver = KNOWN_ROOM_VERSIONS[event_payload["room_version"]]
+                internal_metadata = event_payload["internal_metadata"]
+                rejected_reason = event_payload["rejected_reason"]
+
+                event = make_event_from_dict(
+                    event_dict, room_ver, internal_metadata, rejected_reason
+                )
+                event.internal_metadata.outlier = event_payload["outlier"]
+
+                requester = Requester.deserialize(
+                    self.store, event_payload["requester"]
+                )
+                context = EventContext.deserialize(
+                    self._storage_controllers, event_payload["context"]
+                )
+
+                ratelimit = event_payload["ratelimit"]
+                events_and_context.append((event, context))
+
+                extra_users = [
+                    UserID.from_string(u) for u in event_payload["extra_users"]
+                ]
+
+                logger.info(
+                    "Got batch of events to send, last ID of batch is: %s, sending into room: %s",
+                    event.event_id,
+                    event.room_id,
+                )
+
+            last_event = (
+                await self.event_creation_handler.persist_and_notify_client_events(
+                    requester, events_and_context, ratelimit, extra_users
+                )
+            )
+
+        return (
+            200,
+            {
+                "stream_id": last_event.internal_metadata.stream_ordering,
+                "event_id": last_event.event_id,
+            },
+        )
+
+
+def register_servlets(hs: "HomeServer", http_server: HttpServer) -> None:
+    ReplicationSendEventsRestServlet(hs).register(http_server)
diff --git a/synapse/rest/admin/users.py b/synapse/rest/admin/users.py
index 1274773d7e..15ac2059aa 100644
--- a/synapse/rest/admin/users.py
+++ b/synapse/rest/admin/users.py
@@ -69,6 +69,7 @@ class UsersRestServletV2(RestServlet):
         self.store = hs.get_datastores().main
         self.auth = hs.get_auth()
         self.admin_handler = hs.get_admin_handler()
+        self._msc3866_enabled = hs.config.experimental.msc3866.enabled
 
     async def on_GET(self, request: SynapseRequest) -> Tuple[int, JsonDict]:
         await assert_requester_is_admin(self.auth, request)
@@ -95,6 +96,13 @@ class UsersRestServletV2(RestServlet):
         guests = parse_boolean(request, "guests", default=True)
         deactivated = parse_boolean(request, "deactivated", default=False)
 
+        # If support for MSC3866 is not enabled, apply no filtering based on the
+        # `approved` column.
+        if self._msc3866_enabled:
+            approved = parse_boolean(request, "approved", default=True)
+        else:
+            approved = True
+
         order_by = parse_string(
             request,
             "order_by",
@@ -115,8 +123,22 @@ class UsersRestServletV2(RestServlet):
         direction = parse_string(request, "dir", default="f", allowed_values=("f", "b"))
 
         users, total = await self.store.get_users_paginate(
-            start, limit, user_id, name, guests, deactivated, order_by, direction
+            start,
+            limit,
+            user_id,
+            name,
+            guests,
+            deactivated,
+            order_by,
+            direction,
+            approved,
         )
+
+        # If support for MSC3866 is not enabled, don't show the approval flag.
+        if not self._msc3866_enabled:
+            for user in users:
+                del user["approved"]
+
         ret = {"users": users, "total": total}
         if (start + limit) < total:
             ret["next_token"] = str(start + len(users))
@@ -163,6 +185,7 @@ class UserRestServletV2(RestServlet):
         self.deactivate_account_handler = hs.get_deactivate_account_handler()
         self.registration_handler = hs.get_registration_handler()
         self.pusher_pool = hs.get_pusherpool()
+        self._msc3866_enabled = hs.config.experimental.msc3866.enabled
 
     async def on_GET(
         self, request: SynapseRequest, user_id: str
@@ -239,6 +262,15 @@ class UserRestServletV2(RestServlet):
                 HTTPStatus.BAD_REQUEST, "'deactivated' parameter is not of type boolean"
             )
 
+        approved: Optional[bool] = None
+        if "approved" in body and self._msc3866_enabled:
+            approved = body["approved"]
+            if not isinstance(approved, bool):
+                raise SynapseError(
+                    HTTPStatus.BAD_REQUEST,
+                    "'approved' parameter is not of type boolean",
+                )
+
         # convert List[Dict[str, str]] into List[Tuple[str, str]]
         if external_ids is not None:
             new_external_ids = [
@@ -343,6 +375,9 @@ class UserRestServletV2(RestServlet):
             if "user_type" in body:
                 await self.store.set_user_type(target_user, user_type)
 
+            if approved is not None:
+                await self.store.update_user_approval_status(target_user, approved)
+
             user = await self.admin_handler.get_user(target_user)
             assert user is not None
 
@@ -355,6 +390,10 @@ class UserRestServletV2(RestServlet):
             if password is not None:
                 password_hash = await self.auth_handler.hash(password)
 
+            new_user_approved = True
+            if self._msc3866_enabled and approved is not None:
+                new_user_approved = approved
+
             user_id = await self.registration_handler.register_user(
                 localpart=target_user.localpart,
                 password_hash=password_hash,
@@ -362,6 +401,7 @@ class UserRestServletV2(RestServlet):
                 default_display_name=displayname,
                 user_type=user_type,
                 by_admin=True,
+                approved=new_user_approved,
             )
 
             if threepids is not None:
@@ -550,6 +590,7 @@ class UserRegisterServlet(RestServlet):
             user_type=user_type,
             default_display_name=displayname,
             by_admin=True,
+            approved=True,
         )
 
         result = await register._create_registration_details(user_id, body)
diff --git a/synapse/rest/client/login.py b/synapse/rest/client/login.py
index 0437c87d8d..f554586ac3 100644
--- a/synapse/rest/client/login.py
+++ b/synapse/rest/client/login.py
@@ -28,7 +28,14 @@ from typing import (
 
 from typing_extensions import TypedDict
 
-from synapse.api.errors import Codes, InvalidClientTokenError, LoginError, SynapseError
+from synapse.api.constants import ApprovalNoticeMedium
+from synapse.api.errors import (
+    Codes,
+    InvalidClientTokenError,
+    LoginError,
+    NotApprovedError,
+    SynapseError,
+)
 from synapse.api.ratelimiting import Ratelimiter
 from synapse.api.urls import CLIENT_API_PREFIX
 from synapse.appservice import ApplicationService
@@ -55,11 +62,11 @@ logger = logging.getLogger(__name__)
 
 class LoginResponse(TypedDict, total=False):
     user_id: str
-    access_token: str
+    access_token: Optional[str]
     home_server: str
     expires_in_ms: Optional[int]
     refresh_token: Optional[str]
-    device_id: str
+    device_id: Optional[str]
     well_known: Optional[Dict[str, Any]]
 
 
@@ -92,6 +99,12 @@ class LoginRestServlet(RestServlet):
             hs.config.registration.refreshable_access_token_lifetime is not None
         )
 
+        # Whether we need to check if the user has been approved or not.
+        self._require_approval = (
+            hs.config.experimental.msc3866.enabled
+            and hs.config.experimental.msc3866.require_approval_for_new_accounts
+        )
+
         self.auth = hs.get_auth()
 
         self.clock = hs.get_clock()
@@ -220,6 +233,14 @@ class LoginRestServlet(RestServlet):
         except KeyError:
             raise SynapseError(400, "Missing JSON keys.")
 
+        if self._require_approval:
+            approved = await self.auth_handler.is_user_approved(result["user_id"])
+            if not approved:
+                raise NotApprovedError(
+                    msg="This account is pending approval by a server administrator.",
+                    approval_notice_medium=ApprovalNoticeMedium.NONE,
+                )
+
         well_known_data = self._well_known_builder.get_well_known()
         if well_known_data:
             result["well_known"] = well_known_data
@@ -356,6 +377,16 @@ class LoginRestServlet(RestServlet):
                 errcode=Codes.INVALID_PARAM,
             )
 
+        if self._require_approval:
+            approved = await self.auth_handler.is_user_approved(user_id)
+            if not approved:
+                # If the user isn't approved (and needs to be) we won't allow them to
+                # actually log in, so we don't want to create a device/access token.
+                return LoginResponse(
+                    user_id=user_id,
+                    home_server=self.hs.hostname,
+                )
+
         initial_display_name = login_submission.get("initial_device_display_name")
         (
             device_id,
diff --git a/synapse/rest/client/login_token_request.py b/synapse/rest/client/login_token_request.py
index ca5c54bf17..277b20fb63 100644
--- a/synapse/rest/client/login_token_request.py
+++ b/synapse/rest/client/login_token_request.py
@@ -47,7 +47,9 @@ class LoginTokenRequestServlet(RestServlet):
     }
     """
 
-    PATTERNS = client_patterns("/login/token$")
+    PATTERNS = client_patterns(
+        "/org.matrix.msc3882/login/token$", releases=[], v1=False, unstable=True
+    )
 
     def __init__(self, hs: "HomeServer"):
         super().__init__()
diff --git a/synapse/rest/client/register.py b/synapse/rest/client/register.py
index 20bab20c8f..de810ae3ec 100644
--- a/synapse/rest/client/register.py
+++ b/synapse/rest/client/register.py
@@ -21,10 +21,15 @@ from twisted.web.server import Request
 import synapse
 import synapse.api.auth
 import synapse.types
-from synapse.api.constants import APP_SERVICE_REGISTRATION_TYPE, LoginType
+from synapse.api.constants import (
+    APP_SERVICE_REGISTRATION_TYPE,
+    ApprovalNoticeMedium,
+    LoginType,
+)
 from synapse.api.errors import (
     Codes,
     InteractiveAuthIncompleteError,
+    NotApprovedError,
     SynapseError,
     ThreepidValidationError,
     UnrecognizedRequestError,
@@ -414,6 +419,11 @@ class RegisterRestServlet(RestServlet):
             hs.config.registration.inhibit_user_in_use_error
         )
 
+        self._require_approval = (
+            hs.config.experimental.msc3866.enabled
+            and hs.config.experimental.msc3866.require_approval_for_new_accounts
+        )
+
         self._registration_flows = _calculate_registration_flows(
             hs.config, self.auth_handler
         )
@@ -734,6 +744,12 @@ class RegisterRestServlet(RestServlet):
                 access_token=return_dict.get("access_token"),
             )
 
+            if self._require_approval:
+                raise NotApprovedError(
+                    msg="This account needs to be approved by an administrator before it can be used.",
+                    approval_notice_medium=ApprovalNoticeMedium.NONE,
+                )
+
         return 200, return_dict
 
     async def _do_appservice_registration(
@@ -778,7 +794,9 @@ class RegisterRestServlet(RestServlet):
             "user_id": user_id,
             "home_server": self.hs.hostname,
         }
-        if not params.get("inhibit_login", False):
+        # We don't want to log the user in if we're going to deny them access because
+        # they need to be approved first.
+        if not params.get("inhibit_login", False) and not self._require_approval:
             device_id = params.get("device_id")
             initial_display_name = params.get("initial_device_display_name")
             (
diff --git a/synapse/rest/client/relations.py b/synapse/rest/client/relations.py
index ce97080013..205c556f64 100644
--- a/synapse/rest/client/relations.py
+++ b/synapse/rest/client/relations.py
@@ -56,15 +56,21 @@ class RelationPaginationServlet(RestServlet):
         requester = await self.auth.get_user_by_req(request, allow_guest=True)
 
         limit = parse_integer(request, "limit", default=5)
-        if self._msc3715_enabled:
-            direction = parse_string(
-                request,
-                "org.matrix.msc3715.dir",
-                default="b",
-                allowed_values=["f", "b"],
-            )
-        else:
-            direction = "b"
+        # Fetch the direction parameter, if provided.
+        #
+        # TODO Use PaginationConfig.from_request when the unstable parameter is
+        #      no longer needed.
+        direction = parse_string(request, "dir", allowed_values=["f", "b"])
+        if direction is None:
+            if self._msc3715_enabled:
+                direction = parse_string(
+                    request,
+                    "org.matrix.msc3715.dir",
+                    default="b",
+                    allowed_values=["f", "b"],
+                )
+            else:
+                direction = "b"
         from_token_str = parse_string(request, "from")
         to_token_str = parse_string(request, "to")
 
diff --git a/synapse/state/__init__.py b/synapse/state/__init__.py
index 3787d35b24..6f3dd0463e 100644
--- a/synapse/state/__init__.py
+++ b/synapse/state/__init__.py
@@ -420,6 +420,69 @@ class StateHandler:
             partial_state=partial_state,
         )
 
+    async def compute_event_context_for_batched(
+        self,
+        event: EventBase,
+        state_ids_before_event: StateMap[str],
+        current_state_group: int,
+    ) -> EventContext:
+        """
+        Generate an event context for an event that has not yet been persisted to the
+        database. Intended for use with events that are created to be persisted in a batch.
+        Args:
+            event: the event the context is being computed for
+            state_ids_before_event: a state map consisting of the state ids of the events
+            created prior to this event.
+            current_state_group: the current state group before the event.
+        """
+        state_group_before_event_prev_group = None
+        deltas_to_state_group_before_event = None
+
+        state_group_before_event = current_state_group
+
+        # if the event is not state, we are set
+        if not event.is_state():
+            return EventContext.with_state(
+                storage=self._storage_controllers,
+                state_group_before_event=state_group_before_event,
+                state_group=state_group_before_event,
+                state_delta_due_to_event={},
+                prev_group=state_group_before_event_prev_group,
+                delta_ids=deltas_to_state_group_before_event,
+                partial_state=False,
+            )
+
+        # otherwise, we'll need to create a new state group for after the event
+        key = (event.type, event.state_key)
+
+        if state_ids_before_event is not None:
+            replaces = state_ids_before_event.get(key)
+
+        if replaces and replaces != event.event_id:
+            event.unsigned["replaces_state"] = replaces
+
+        delta_ids = {key: event.event_id}
+
+        state_group_after_event = (
+            await self._state_storage_controller.store_state_group(
+                event.event_id,
+                event.room_id,
+                prev_group=state_group_before_event,
+                delta_ids=delta_ids,
+                current_state_ids=None,
+            )
+        )
+
+        return EventContext.with_state(
+            storage=self._storage_controllers,
+            state_group=state_group_after_event,
+            state_group_before_event=state_group_before_event,
+            state_delta_due_to_event=delta_ids,
+            prev_group=state_group_before_event,
+            delta_ids=delta_ids,
+            partial_state=False,
+        )
+
     @measure_func()
     async def resolve_state_groups_for_events(
         self, room_id: str, event_ids: Collection[str], await_full_state: bool = True
diff --git a/synapse/storage/_base.py b/synapse/storage/_base.py
index 313e8aca7d..bf42aeb8d1 100644
--- a/synapse/storage/_base.py
+++ b/synapse/storage/_base.py
@@ -94,6 +94,7 @@ class SQLBaseStore(metaclass=ABCMeta):
             self._attempt_to_invalidate_cache(
                 "get_rooms_for_user_with_stream_ordering", (user_id,)
             )
+            self._attempt_to_invalidate_cache("get_rooms_for_user", (user_id,))
 
         # Purge other caches based on room state.
         self._attempt_to_invalidate_cache("get_room_summary", (room_id,))
diff --git a/synapse/storage/controllers/persist_events.py b/synapse/storage/controllers/persist_events.py
index bde7a6648a..e18c7fb85d 100644
--- a/synapse/storage/controllers/persist_events.py
+++ b/synapse/storage/controllers/persist_events.py
@@ -427,16 +427,18 @@ class EventsPersistenceStorageController:
         for d in ret_vals:
             replaced_events.update(d)
 
-        events = []
+        persisted_events = []
         for event, _ in events_and_contexts:
             existing_event_id = replaced_events.get(event.event_id)
             if existing_event_id:
-                events.append(await self.main_store.get_event(existing_event_id))
+                persisted_events.append(
+                    await self.main_store.get_event(existing_event_id)
+                )
             else:
-                events.append(event)
+                persisted_events.append(event)
 
         return (
-            events,
+            persisted_events,
             self.main_store.get_room_max_token(),
         )
 
@@ -602,11 +604,6 @@ class EventsPersistenceStorageController:
             # room
             state_delta_for_room: Dict[str, DeltaState] = {}
 
-            # Set of remote users which were in rooms the server has left or who may
-            # have left rooms the server is in. We should check if we still share any
-            # rooms and if not we mark their device lists as stale.
-            potentially_left_users: Set[str] = set()
-
             if not backfilled:
                 with Measure(self._clock, "_calculate_state_and_extrem"):
                     # Work out the new "current state" for each room.
@@ -720,8 +717,6 @@ class EventsPersistenceStorageController:
                                 room_id,
                                 ev_ctx_rm,
                                 delta,
-                                current_state,
-                                potentially_left_users,
                             )
                             if not is_still_joined:
                                 logger.info("Server no longer in room %s", room_id)
@@ -729,20 +724,6 @@ class EventsPersistenceStorageController:
                                 current_state = {}
                                 delta.no_longer_in_room = True
 
-                            # Add all remote users that might have left rooms.
-                            potentially_left_users.update(
-                                user_id
-                                for event_type, user_id in delta.to_delete
-                                if event_type == EventTypes.Member
-                                and not self.is_mine_id(user_id)
-                            )
-                            potentially_left_users.update(
-                                user_id
-                                for event_type, user_id in delta.to_insert.keys()
-                                if event_type == EventTypes.Member
-                                and not self.is_mine_id(user_id)
-                            )
-
                             state_delta_for_room[room_id] = delta
 
             await self.persist_events_store._persist_events_and_state_updates(
@@ -753,8 +734,6 @@ class EventsPersistenceStorageController:
                 inhibit_local_membership_updates=backfilled,
             )
 
-            await self._handle_potentially_left_users(potentially_left_users)
-
         return replaced_events
 
     async def _calculate_new_extremities(
@@ -1130,8 +1109,6 @@ class EventsPersistenceStorageController:
         room_id: str,
         ev_ctx_rm: List[Tuple[EventBase, EventContext]],
         delta: DeltaState,
-        current_state: Optional[StateMap[str]],
-        potentially_left_users: Set[str],
     ) -> bool:
         """Check if the server will still be joined after the given events have
         been persised.
@@ -1141,11 +1118,6 @@ class EventsPersistenceStorageController:
             ev_ctx_rm
             delta: The delta of current state between what is in the database
                 and what the new current state will be.
-            current_state: The new current state if it already been calculated,
-                otherwise None.
-            potentially_left_users: If the server has left the room, then joined
-                remote users will be added to this set to indicate that the
-                server may no longer be sharing a room with them.
         """
 
         if not any(
@@ -1199,45 +1171,4 @@ class EventsPersistenceStorageController:
         ):
             return True
 
-        # The server will leave the room, so we go and find out which remote
-        # users will still be joined when we leave.
-        if current_state is None:
-            current_state = await self.main_store.get_partial_current_state_ids(room_id)
-            current_state = dict(current_state)
-            for key in delta.to_delete:
-                current_state.pop(key, None)
-
-            current_state.update(delta.to_insert)
-
-        remote_event_ids = [
-            event_id
-            for (
-                typ,
-                state_key,
-            ), event_id in current_state.items()
-            if typ == EventTypes.Member and not self.is_mine_id(state_key)
-        ]
-        members = await self.main_store.get_membership_from_event_ids(remote_event_ids)
-        potentially_left_users.update(
-            member.user_id
-            for member in members.values()
-            if member and member.membership == Membership.JOIN
-        )
-
         return False
-
-    async def _handle_potentially_left_users(self, user_ids: Set[str]) -> None:
-        """Given a set of remote users check if the server still shares a room with
-        them. If not then mark those users' device cache as stale.
-        """
-
-        if not user_ids:
-            return
-
-        joined_users = await self.main_store.get_users_server_still_shares_room_with(
-            user_ids
-        )
-        left_users = user_ids - joined_users
-
-        for user_id in left_users:
-            await self.main_store.mark_remote_user_device_list_as_unsubscribed(user_id)
diff --git a/synapse/storage/controllers/state.py b/synapse/storage/controllers/state.py
index b4f055fb60..263b26e345 100644
--- a/synapse/storage/controllers/state.py
+++ b/synapse/storage/controllers/state.py
@@ -23,7 +23,7 @@ from typing import (
     List,
     Mapping,
     Optional,
-    Sequence,
+    Set,
     Tuple,
 )
 
@@ -529,7 +529,18 @@ class StateStorageController:
         )
         return state_map.get(key)
 
-    async def get_current_hosts_in_room(self, room_id: str) -> List[str]:
+    async def get_current_hosts_in_room(self, room_id: str) -> Set[str]:
+        """Get current hosts in room based on current state.
+
+        Blocks until we have full state for the given room. This only happens for rooms
+        with partial state.
+        """
+
+        await self._partial_state_room_tracker.await_full_state(room_id)
+
+        return await self.stores.main.get_current_hosts_in_room(room_id)
+
+    async def get_current_hosts_in_room_ordered(self, room_id: str) -> List[str]:
         """Get current hosts in room based on current state.
 
         Blocks until we have full state for the given room. This only happens for rooms
@@ -542,11 +553,11 @@ class StateStorageController:
 
         await self._partial_state_room_tracker.await_full_state(room_id)
 
-        return await self.stores.main.get_current_hosts_in_room(room_id)
+        return await self.stores.main.get_current_hosts_in_room_ordered(room_id)
 
     async def get_current_hosts_in_room_or_partial_state_approximation(
         self, room_id: str
-    ) -> Sequence[str]:
+    ) -> Collection[str]:
         """Get approximation of current hosts in room based on current state.
 
         For rooms with full state, this is equivalent to `get_current_hosts_in_room`,
@@ -566,14 +577,9 @@ class StateStorageController:
         )
 
         hosts_from_state = await self.stores.main.get_current_hosts_in_room(room_id)
-        hosts_from_state_set = set(hosts_from_state)
-
-        # First take the list of hosts based on the current state.
-        # For rooms with partial state, this will be missing most hosts.
-        hosts = list(hosts_from_state)
-        # Then add in the list of hosts in the room at the time we joined.
-        # This will be an empty list for rooms with full state.
-        hosts.extend(host for host in hosts_at_join if host not in hosts_from_state_set)
+
+        hosts = set(hosts_at_join)
+        hosts.update(hosts_from_state)
 
         return hosts
 
diff --git a/synapse/storage/database.py b/synapse/storage/database.py
index a00f8dcd21..3edeed7fd8 100644
--- a/synapse/storage/database.py
+++ b/synapse/storage/database.py
@@ -290,8 +290,7 @@ class LoggingTransaction:
         # LoggingTransaction isn't expecting there to be any callbacks; assert that
         # is not the case.
         assert self.after_callbacks is not None
-        # type-ignore: need mypy containing https://github.com/python/mypy/pull/12668
-        self.after_callbacks.append((callback, args, kwargs))  # type: ignore[arg-type]
+        self.after_callbacks.append((callback, args, kwargs))
 
     def async_call_after(
         self, callback: Callable[P, Awaitable], *args: P.args, **kwargs: P.kwargs
@@ -312,8 +311,7 @@ class LoggingTransaction:
         # LoggingTransaction isn't expecting there to be any callbacks; assert that
         # is not the case.
         assert self.async_after_callbacks is not None
-        # type-ignore: need mypy containing https://github.com/python/mypy/pull/12668
-        self.async_after_callbacks.append((callback, args, kwargs))  # type: ignore[arg-type]
+        self.async_after_callbacks.append((callback, args, kwargs))
 
     def call_on_exception(
         self, callback: Callable[P, object], *args: P.args, **kwargs: P.kwargs
@@ -331,8 +329,7 @@ class LoggingTransaction:
         # LoggingTransaction isn't expecting there to be any callbacks; assert that
         # is not the case.
         assert self.exception_callbacks is not None
-        # type-ignore: need mypy containing https://github.com/python/mypy/pull/12668
-        self.exception_callbacks.append((callback, args, kwargs))  # type: ignore[arg-type]
+        self.exception_callbacks.append((callback, args, kwargs))
 
     def fetchone(self) -> Optional[Tuple]:
         return self.txn.fetchone()
@@ -421,10 +418,7 @@ class LoggingTransaction:
         sql = self.database_engine.convert_param_style(sql)
         if args:
             try:
-                # The type-ignore should be redundant once mypy releases a version with
-                # https://github.com/python/mypy/pull/12668. (`args` might be empty,
-                # (but we'll catch the index error if so.)
-                sql_logger.debug("[SQL values] {%s} %r", self.name, args[0])  # type: ignore[index]
+                sql_logger.debug("[SQL values] {%s} %r", self.name, args[0])
             except Exception:
                 # Don't let logging failures stop SQL from working
                 pass
@@ -655,9 +649,7 @@ class DatabasePool:
         # For now, we just log an error, and hope that it works on the first attempt.
         # TODO: raise an exception.
 
-        # Type-ignore Mypy doesn't yet consider ParamSpec.args to be iterable; see
-        # https://github.com/python/mypy/pull/12668
-        for i, arg in enumerate(args):  # type: ignore[arg-type, var-annotated]
+        for i, arg in enumerate(args):
             if inspect.isgenerator(arg):
                 logger.error(
                     "Programming error: generator passed to new_transaction as "
@@ -665,9 +657,7 @@ class DatabasePool:
                     i,
                     func,
                 )
-        # Type-ignore Mypy doesn't yet consider ParamSpec.args to be a mapping; see
-        # https://github.com/python/mypy/pull/12668
-        for name, val in kwargs.items():  # type: ignore[attr-defined]
+        for name, val in kwargs.items():
             if inspect.isgenerator(val):
                 logger.error(
                     "Programming error: generator passed to new_transaction as "
@@ -1137,17 +1127,57 @@ class DatabasePool:
         desc: str = "simple_upsert",
         lock: bool = True,
     ) -> bool:
-        """
+        """Insert a row with values + insertion_values; on conflict, update with values.
+
+        All of our supported databases accept the nonstandard "upsert" statement in
+        their dialect of SQL. We call this a "native upsert". The syntax looks roughly
+        like:
+
+            INSERT INTO table VALUES (values + insertion_values)
+            ON CONFLICT (keyvalues)
+            DO UPDATE SET (values); -- overwrite `values` columns only
+
+        If (values) is empty, the resulting query is slighlty simpler:
+
+            INSERT INTO table VALUES (insertion_values)
+            ON CONFLICT (keyvalues)
+            DO NOTHING;             -- do not overwrite any columns
+
+        This function is a helper to build such queries.
+
+        In order for upserts to make sense, the database must be able to determine when
+        an upsert CONFLICTs with an existing row. Postgres and SQLite ensure this by
+        requiring that a unique index exist on the column names used to detect a
+        conflict (i.e. `keyvalues.keys()`).
+
+        If there is no such index, we can "emulate" an upsert with a SELECT followed
+        by either an INSERT or an UPDATE. This is unsafe: we cannot make the same
+        atomicity guarantees that a native upsert can and are very vulnerable to races
+        and crashes. Therefore if we wish to upsert without an appropriate unique index,
+        we must either:
+
+        1. Acquire a table-level lock before the emulated upsert (`lock=True`), or
+        2. VERY CAREFULLY ensure that we are the only thread and worker which will be
+           writing to this table, in which case we can proceed without a lock
+           (`lock=False`).
+
+        Generally speaking, you should use `lock=True`. If the table in question has a
+        unique index[*], this class will use a native upsert (which is atomic and so can
+        ignore the `lock` argument). Otherwise this class will use an emulated upsert,
+        in which case we want the safer option unless we been VERY CAREFUL.
+
+        [*]: Some tables have unique indices added to them in the background. Those
+             tables `T` are keys in the dictionary UNIQUE_INDEX_BACKGROUND_UPDATES,
+             where `T` maps to the background update that adds a unique index to `T`.
+             This dictionary is maintained by hand.
+
+             At runtime, we constantly check to see if each of these background updates
+             has run. If so, we deem the coresponding table safe to upsert into, because
+             we can now use a native insert to do so. If not, we deem the table unsafe
+             to upsert into and require an emulated upsert.
 
-        `lock` should generally be set to True (the default), but can be set
-        to False if either of the following are true:
-            1. there is a UNIQUE INDEX on the key columns. In this case a conflict
-            will cause an IntegrityError in which case this function will retry
-            the update.
-            2. we somehow know that we are the only thread which will be updating
-            this table.
-        As an additional note, this parameter only matters for old SQLite versions
-        because we will use native upserts otherwise.
+             Tables that do not appear in this dictionary are assumed to have an
+             appropriate unique index and therefore be safe to upsert into.
 
         Args:
             table: The table to upsert into
diff --git a/synapse/storage/databases/main/__init__.py b/synapse/storage/databases/main/__init__.py
index 4dccbb732a..a62b4abd4e 100644
--- a/synapse/storage/databases/main/__init__.py
+++ b/synapse/storage/databases/main/__init__.py
@@ -83,6 +83,7 @@ logger = logging.getLogger(__name__)
 
 class DataStore(
     EventsBackgroundUpdatesStore,
+    DeviceStore,
     RoomMemberStore,
     RoomStore,
     RoomBatchStore,
@@ -114,7 +115,6 @@ class DataStore(
     StreamWorkerStore,
     OpenIdStore,
     ClientIpWorkerStore,
-    DeviceStore,
     DeviceInboxStore,
     UserDirectoryStore,
     UserErasureStore,
@@ -203,6 +203,7 @@ class DataStore(
         deactivated: bool = False,
         order_by: str = UserSortOrder.USER_ID.value,
         direction: str = "f",
+        approved: bool = True,
     ) -> Tuple[List[JsonDict], int]:
         """Function to retrieve a paginated list of users from
         users list. This will return a json list of users and the
@@ -217,6 +218,7 @@ class DataStore(
             deactivated: whether to include deactivated users
             order_by: the sort order of the returned list
             direction: sort ascending or descending
+            approved: whether to include approved users
         Returns:
             A tuple of a list of mappings from user to information and a count of total users.
         """
@@ -249,6 +251,11 @@ class DataStore(
             if not deactivated:
                 filters.append("deactivated = 0")
 
+            if not approved:
+                # We ignore NULL values for the approved flag because these should only
+                # be already existing users that we consider as already approved.
+                filters.append("approved IS FALSE")
+
             where_clause = "WHERE " + " AND ".join(filters) if len(filters) > 0 else ""
 
             sql_base = f"""
@@ -262,7 +269,7 @@ class DataStore(
 
             sql = f"""
                 SELECT name, user_type, is_guest, admin, deactivated, shadow_banned,
-                displayname, avatar_url, creation_ts * 1000 as creation_ts
+                displayname, avatar_url, creation_ts * 1000 as creation_ts, approved
                 {sql_base}
                 ORDER BY {order_by_column} {order}, u.name ASC
                 LIMIT ? OFFSET ?
diff --git a/synapse/storage/databases/main/cache.py b/synapse/storage/databases/main/cache.py
index db6ce83a2b..3b8ed1f7ee 100644
--- a/synapse/storage/databases/main/cache.py
+++ b/synapse/storage/databases/main/cache.py
@@ -205,6 +205,7 @@ class CacheInvalidationWorkerStore(SQLBaseStore):
                 self.get_rooms_for_user_with_stream_ordering.invalidate(
                     (data.state_key,)
                 )
+                self.get_rooms_for_user.invalidate((data.state_key,))
         else:
             raise Exception("Unknown events stream row type %s" % (row.type,))
 
diff --git a/synapse/storage/databases/main/devices.py b/synapse/storage/databases/main/devices.py
index dfca34550d..d7f015c783 100644
--- a/synapse/storage/databases/main/devices.py
+++ b/synapse/storage/databases/main/devices.py
@@ -47,6 +47,7 @@ from synapse.storage.database import (
     make_tuple_comparison_clause,
 )
 from synapse.storage.databases.main.end_to_end_keys import EndToEndKeyWorkerStore
+from synapse.storage.databases.main.roommember import RoomMemberWorkerStore
 from synapse.storage.types import Cursor
 from synapse.types import JsonDict, get_verify_key_from_cross_signing_key
 from synapse.util import json_decoder, json_encoder
@@ -70,7 +71,7 @@ DROP_DEVICE_LIST_STREAMS_NON_UNIQUE_INDEXES = (
 BG_UPDATE_REMOVE_DUP_OUTBOUND_POKES = "remove_dup_outbound_pokes"
 
 
-class DeviceWorkerStore(EndToEndKeyWorkerStore):
+class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore):
     def __init__(
         self,
         database: DatabasePool,
@@ -985,24 +986,59 @@ class DeviceWorkerStore(EndToEndKeyWorkerStore):
             desc="mark_remote_user_device_cache_as_valid",
         )
 
+    async def handle_potentially_left_users(self, user_ids: Set[str]) -> None:
+        """Given a set of remote users check if the server still shares a room with
+        them. If not then mark those users' device cache as stale.
+        """
+
+        if not user_ids:
+            return
+
+        await self.db_pool.runInteraction(
+            "_handle_potentially_left_users",
+            self.handle_potentially_left_users_txn,
+            user_ids,
+        )
+
+    def handle_potentially_left_users_txn(
+        self,
+        txn: LoggingTransaction,
+        user_ids: Set[str],
+    ) -> None:
+        """Given a set of remote users check if the server still shares a room with
+        them. If not then mark those users' device cache as stale.
+        """
+
+        if not user_ids:
+            return
+
+        joined_users = self.get_users_server_still_shares_room_with_txn(txn, user_ids)
+        left_users = user_ids - joined_users
+
+        for user_id in left_users:
+            self.mark_remote_user_device_list_as_unsubscribed_txn(txn, user_id)
+
     async def mark_remote_user_device_list_as_unsubscribed(self, user_id: str) -> None:
         """Mark that we no longer track device lists for remote user."""
 
-        def _mark_remote_user_device_list_as_unsubscribed_txn(
-            txn: LoggingTransaction,
-        ) -> None:
-            self.db_pool.simple_delete_txn(
-                txn,
-                table="device_lists_remote_extremeties",
-                keyvalues={"user_id": user_id},
-            )
-            self._invalidate_cache_and_stream(
-                txn, self.get_device_list_last_stream_id_for_remote, (user_id,)
-            )
-
         await self.db_pool.runInteraction(
             "mark_remote_user_device_list_as_unsubscribed",
-            _mark_remote_user_device_list_as_unsubscribed_txn,
+            self.mark_remote_user_device_list_as_unsubscribed_txn,
+            user_id,
+        )
+
+    def mark_remote_user_device_list_as_unsubscribed_txn(
+        self,
+        txn: LoggingTransaction,
+        user_id: str,
+    ) -> None:
+        self.db_pool.simple_delete_txn(
+            txn,
+            table="device_lists_remote_extremeties",
+            keyvalues={"user_id": user_id},
+        )
+        self._invalidate_cache_and_stream(
+            txn, self.get_device_list_last_stream_id_for_remote, (user_id,)
         )
 
     async def get_dehydrated_device(
@@ -1271,6 +1307,33 @@ class DeviceWorkerStore(EndToEndKeyWorkerStore):
 
         return changes
 
+    async def get_device_list_changes_in_room(
+        self, room_id: str, min_stream_id: int
+    ) -> Collection[Tuple[str, str]]:
+        """Get all device list changes that happened in the room since the given
+        stream ID.
+
+        Returns:
+            Collection of user ID/device ID tuples of all devices that have
+            changed
+        """
+
+        sql = """
+            SELECT DISTINCT user_id, device_id FROM device_lists_changes_in_room
+            WHERE room_id = ? AND stream_id > ?
+        """
+
+        def get_device_list_changes_in_room_txn(
+            txn: LoggingTransaction,
+        ) -> Collection[Tuple[str, str]]:
+            txn.execute(sql, (room_id, min_stream_id))
+            return cast(Collection[Tuple[str, str]], txn.fetchall())
+
+        return await self.db_pool.runInteraction(
+            "get_device_list_changes_in_room",
+            get_device_list_changes_in_room_txn,
+        )
+
 
 class DeviceBackgroundUpdateStore(SQLBaseStore):
     def __init__(
@@ -1910,14 +1973,15 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore):
         user_id: str,
         device_id: str,
         room_id: str,
-        stream_id: int,
+        stream_id: Optional[int],
         hosts: Collection[str],
         context: Optional[Dict[str, str]],
     ) -> None:
         """Queue the device update to be sent to the given set of hosts,
         calculated from the room ID.
 
-        Marks the associated row in `device_lists_changes_in_room` as handled.
+        Marks the associated row in `device_lists_changes_in_room` as handled,
+        if `stream_id` is provided.
         """
 
         def add_device_list_outbound_pokes_txn(
@@ -1933,17 +1997,18 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore):
                     context=context,
                 )
 
-            self.db_pool.simple_update_txn(
-                txn,
-                table="device_lists_changes_in_room",
-                keyvalues={
-                    "user_id": user_id,
-                    "device_id": device_id,
-                    "stream_id": stream_id,
-                    "room_id": room_id,
-                },
-                updatevalues={"converted_to_destinations": True},
-            )
+            if stream_id:
+                self.db_pool.simple_update_txn(
+                    txn,
+                    table="device_lists_changes_in_room",
+                    keyvalues={
+                        "user_id": user_id,
+                        "device_id": device_id,
+                        "stream_id": stream_id,
+                        "room_id": room_id,
+                    },
+                    updatevalues={"converted_to_destinations": True},
+                )
 
         if not hosts:
             # If there are no hosts then we don't try and generate stream IDs.
@@ -1959,3 +2024,58 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore):
                 add_device_list_outbound_pokes_txn,
                 stream_ids,
             )
+
+    async def add_remote_device_list_to_pending(
+        self, user_id: str, device_id: str
+    ) -> None:
+        """Add a device list update to the table tracking remote device list
+        updates during partial joins.
+        """
+
+        async with self._device_list_id_gen.get_next() as stream_id:  # type: ignore[attr-defined]
+            await self.db_pool.simple_upsert(
+                table="device_lists_remote_pending",
+                keyvalues={
+                    "user_id": user_id,
+                    "device_id": device_id,
+                },
+                values={"stream_id": stream_id},
+                desc="add_remote_device_list_to_pending",
+            )
+
+    async def get_pending_remote_device_list_updates_for_room(
+        self, room_id: str
+    ) -> Collection[Tuple[str, str]]:
+        """Get the set of remote device list updates from the pending table for
+        the room.
+        """
+
+        min_device_stream_id = await self.db_pool.simple_select_one_onecol(
+            table="partial_state_rooms",
+            keyvalues={
+                "room_id": room_id,
+            },
+            retcol="device_lists_stream_id",
+            desc="get_pending_remote_device_list_updates_for_room_device",
+        )
+
+        sql = """
+            SELECT user_id, device_id FROM device_lists_remote_pending AS d
+            INNER JOIN current_state_events AS c ON
+                type = 'm.room.member'
+                AND state_key = user_id
+                AND membership = 'join'
+            WHERE
+                room_id = ? AND stream_id > ?
+        """
+
+        def get_pending_remote_device_list_updates_for_room_txn(
+            txn: LoggingTransaction,
+        ) -> Collection[Tuple[str, str]]:
+            txn.execute(sql, (room_id, min_device_stream_id))
+            return cast(Collection[Tuple[str, str]], txn.fetchall())
+
+        return await self.db_pool.runInteraction(
+            "get_pending_remote_device_list_updates_for_room",
+            get_pending_remote_device_list_updates_for_room_txn,
+        )
diff --git a/synapse/storage/databases/main/event_federation.py b/synapse/storage/databases/main/event_federation.py
index 0261ff4ad6..57fd4c4587 100644
--- a/synapse/storage/databases/main/event_federation.py
+++ b/synapse/storage/databases/main/event_federation.py
@@ -73,13 +73,30 @@ pdus_pruned_from_federation_queue = Counter(
 
 logger = logging.getLogger(__name__)
 
-BACKFILL_EVENT_BACKOFF_UPPER_BOUND_SECONDS: int = int(
-    datetime.timedelta(days=7).total_seconds()
-)
-BACKFILL_EVENT_EXPONENTIAL_BACKOFF_STEP_SECONDS: int = int(
-    datetime.timedelta(hours=1).total_seconds()
+# Parameters controlling exponential backoff between backfill failures.
+# After the first failure to backfill, we wait 2 hours before trying again. If the
+# second attempt fails, we wait 4 hours before trying again. If the third attempt fails,
+# we wait 8 hours before trying again, ... and so on.
+#
+# Each successive backoff period is twice as long as the last. However we cap this
+# period at a maximum of 2^8 = 256 hours: a little over 10 days. (This is the smallest
+# power of 2 which yields a maximum backoff period of at least 7 days---which was the
+# original maximum backoff period.) Even when we hit this cap, we will continue to
+# make backfill attempts once every 10 days.
+BACKFILL_EVENT_EXPONENTIAL_BACKOFF_MAXIMUM_DOUBLING_STEPS = 8
+BACKFILL_EVENT_EXPONENTIAL_BACKOFF_STEP_MILLISECONDS = int(
+    datetime.timedelta(hours=1).total_seconds() * 1000
 )
 
+# We need a cap on the power of 2 or else the backoff period
+#   2^N * (milliseconds per hour)
+# will overflow when calcuated within the database. We ensure overflow does not occur
+# by checking that the largest backoff period fits in a 32-bit signed integer.
+_LONGEST_BACKOFF_PERIOD_MILLISECONDS = (
+    2**BACKFILL_EVENT_EXPONENTIAL_BACKOFF_MAXIMUM_DOUBLING_STEPS
+) * BACKFILL_EVENT_EXPONENTIAL_BACKOFF_STEP_MILLISECONDS
+assert 0 < _LONGEST_BACKOFF_PERIOD_MILLISECONDS <= ((2**31) - 1)
+
 
 # All the info we need while iterating the DAG while backfilling
 @attr.s(frozen=True, slots=True, auto_attribs=True)
@@ -726,17 +743,35 @@ class EventFederationWorkerStore(SignatureWorkerStore, EventsWorkerStore, SQLBas
     async def get_backfill_points_in_room(
         self,
         room_id: str,
+        current_depth: int,
+        limit: int,
     ) -> List[Tuple[str, int]]:
         """
-        Gets the oldest events(backwards extremities) in the room along with the
-        approximate depth. Sorted by depth, highest to lowest (descending).
+        Get the backward extremities to backfill from in the room along with the
+        approximate depth.
+
+        Only returns events that are at a depth lower than or
+        equal to the `current_depth`. Sorted by depth, highest to lowest (descending)
+        so the closest events to the `current_depth` are first in the list.
+
+        We ignore extremities that are newer than the user's current scroll position
+        (ie, those with depth greater than `current_depth`) as:
+            1. we don't really care about getting events that have happened
+               after our current position; and
+            2. by the nature of paginating and scrolling back, we have likely
+               previously tried and failed to backfill from that extremity, so
+               to avoid getting "stuck" requesting the same backfill repeatedly
+               we drop those extremities.
 
         Args:
             room_id: Room where we want to find the oldest events
+            current_depth: The depth at the user's current scrollback position
+            limit: The max number of backfill points to return
 
         Returns:
             List of (event_id, depth) tuples. Sorted by depth, highest to lowest
-            (descending)
+            (descending) so the closest events to the `current_depth` are first
+            in the list.
         """
 
         def get_backfill_points_in_room_txn(
@@ -749,7 +784,15 @@ class EventFederationWorkerStore(SignatureWorkerStore, EventsWorkerStore, SQLBas
             # persisted in our database yet (meaning we don't know their depth
             # specifically). So we need to look for the approximate depth from
             # the events connected to the current backwards extremeties.
-            sql = """
+
+            if isinstance(self.database_engine, PostgresEngine):
+                least_function = "LEAST"
+            elif isinstance(self.database_engine, Sqlite3Engine):
+                least_function = "MIN"
+            else:
+                raise RuntimeError("Unknown database engine")
+
+            sql = f"""
                 SELECT backward_extrem.event_id, event.depth FROM events AS event
                 /**
                  * Get the edge connections from the event_edges table
@@ -785,6 +828,18 @@ class EventFederationWorkerStore(SignatureWorkerStore, EventsWorkerStore, SQLBas
                      */
                     AND edge.is_state is ? /* False */
                     /**
+                     * We only want backwards extremities that are older than or at
+                     * the same position of the given `current_depth` (where older
+                     * means less than the given depth) because we're looking backwards
+                     * from the `current_depth` when backfilling.
+                     *
+                     *                         current_depth (ignore events that come after this, ignore 2-4)
+                     *                         |
+                     *                         ▼
+                     * <oldest-in-time> [0]<--[1]<--[2]<--[3]<--[4] <newest-in-time>
+                     */
+                    AND event.depth <= ? /* current_depth */
+                    /**
                      * Exponential back-off (up to the upper bound) so we don't retry the
                      * same backfill point over and over. ex. 2hr, 4hr, 8hr, 16hr, etc.
                      *
@@ -795,31 +850,31 @@ class EventFederationWorkerStore(SignatureWorkerStore, EventsWorkerStore, SQLBas
                      */
                     AND (
                         failed_backfill_attempt_info.event_id IS NULL
-                        OR ? /* current_time */ >= failed_backfill_attempt_info.last_attempt_ts + /*least*/%s((1 << failed_backfill_attempt_info.num_attempts) * ? /* step */, ? /* upper bound */)
+                        OR ? /* current_time */ >= failed_backfill_attempt_info.last_attempt_ts + (
+                            (1 << {least_function}(failed_backfill_attempt_info.num_attempts, ? /* max doubling steps */))
+                            * ? /* step */
+                        )
                     )
                 /**
-                 * Sort from highest to the lowest depth. Then tie-break on
-                 * alphabetical order of the event_ids so we get a consistent
-                 * ordering which is nice when asserting things in tests.
+                 * Sort from highest (closest to the `current_depth`) to the lowest depth
+                 * because the closest are most relevant to backfill from first.
+                 * Then tie-break on alphabetical order of the event_ids so we get a
+                 * consistent ordering which is nice when asserting things in tests.
                  */
                 ORDER BY event.depth DESC, backward_extrem.event_id DESC
+                LIMIT ?
             """
 
-            if isinstance(self.database_engine, PostgresEngine):
-                least_function = "least"
-            elif isinstance(self.database_engine, Sqlite3Engine):
-                least_function = "min"
-            else:
-                raise RuntimeError("Unknown database engine")
-
             txn.execute(
-                sql % (least_function,),
+                sql,
                 (
                     room_id,
                     False,
+                    current_depth,
                     self._clock.time_msec(),
-                    1000 * BACKFILL_EVENT_EXPONENTIAL_BACKOFF_STEP_SECONDS,
-                    1000 * BACKFILL_EVENT_BACKOFF_UPPER_BOUND_SECONDS,
+                    BACKFILL_EVENT_EXPONENTIAL_BACKOFF_MAXIMUM_DOUBLING_STEPS,
+                    BACKFILL_EVENT_EXPONENTIAL_BACKOFF_STEP_MILLISECONDS,
+                    limit,
                 ),
             )
 
@@ -835,24 +890,47 @@ class EventFederationWorkerStore(SignatureWorkerStore, EventsWorkerStore, SQLBas
     async def get_insertion_event_backward_extremities_in_room(
         self,
         room_id: str,
+        current_depth: int,
+        limit: int,
     ) -> List[Tuple[str, int]]:
         """
         Get the insertion events we know about that we haven't backfilled yet
-        along with the approximate depth. Sorted by depth, highest to lowest
-        (descending).
+        along with the approximate depth. Only returns insertion events that are
+        at a depth lower than or equal to the `current_depth`. Sorted by depth,
+        highest to lowest (descending) so the closest events to the
+        `current_depth` are first in the list.
+
+        We ignore insertion events that are newer than the user's current scroll
+        position (ie, those with depth greater than `current_depth`) as:
+            1. we don't really care about getting events that have happened
+               after our current position; and
+            2. by the nature of paginating and scrolling back, we have likely
+               previously tried and failed to backfill from that insertion event, so
+               to avoid getting "stuck" requesting the same backfill repeatedly
+               we drop those insertion event.
 
         Args:
             room_id: Room where we want to find the oldest events
+            current_depth: The depth at the user's current scrollback position
+            limit: The max number of insertion event extremities to return
 
         Returns:
             List of (event_id, depth) tuples. Sorted by depth, highest to lowest
-            (descending)
+            (descending) so the closest events to the `current_depth` are first
+            in the list.
         """
 
         def get_insertion_event_backward_extremities_in_room_txn(
             txn: LoggingTransaction, room_id: str
         ) -> List[Tuple[str, int]]:
-            sql = """
+            if isinstance(self.database_engine, PostgresEngine):
+                least_function = "LEAST"
+            elif isinstance(self.database_engine, Sqlite3Engine):
+                least_function = "MIN"
+            else:
+                raise RuntimeError("Unknown database engine")
+
+            sql = f"""
                 SELECT
                     insertion_event_extremity.event_id, event.depth
                 /* We only want insertion events that are also marked as backwards extremities */
@@ -870,6 +948,18 @@ class EventFederationWorkerStore(SignatureWorkerStore, EventsWorkerStore, SQLBas
                 WHERE
                     insertion_event_extremity.room_id = ?
                     /**
+                     * We only want extremities that are older than or at
+                     * the same position of the given `current_depth` (where older
+                     * means less than the given depth) because we're looking backwards
+                     * from the `current_depth` when backfilling.
+                     *
+                     *                         current_depth (ignore events that come after this, ignore 2-4)
+                     *                         |
+                     *                         ▼
+                     * <oldest-in-time> [0]<--[1]<--[2]<--[3]<--[4] <newest-in-time>
+                     */
+                    AND event.depth <= ? /* current_depth */
+                    /**
                      * Exponential back-off (up to the upper bound) so we don't retry the
                      * same backfill point over and over. ex. 2hr, 4hr, 8hr, 16hr, etc
                      *
@@ -880,30 +970,30 @@ class EventFederationWorkerStore(SignatureWorkerStore, EventsWorkerStore, SQLBas
                      */
                     AND (
                         failed_backfill_attempt_info.event_id IS NULL
-                        OR ? /* current_time */ >= failed_backfill_attempt_info.last_attempt_ts + /*least*/%s((1 << failed_backfill_attempt_info.num_attempts) * ? /* step */, ? /* upper bound */)
+                        OR ? /* current_time */ >= failed_backfill_attempt_info.last_attempt_ts + (
+                            (1 << {least_function}(failed_backfill_attempt_info.num_attempts, ? /* max doubling steps */))
+                            * ? /* step */
+                        )
                     )
                 /**
-                 * Sort from highest to the lowest depth. Then tie-break on
-                 * alphabetical order of the event_ids so we get a consistent
-                 * ordering which is nice when asserting things in tests.
+                 * Sort from highest (closest to the `current_depth`) to the lowest depth
+                 * because the closest are most relevant to backfill from first.
+                 * Then tie-break on alphabetical order of the event_ids so we get a
+                 * consistent ordering which is nice when asserting things in tests.
                  */
                 ORDER BY event.depth DESC, insertion_event_extremity.event_id DESC
+                LIMIT ?
             """
 
-            if isinstance(self.database_engine, PostgresEngine):
-                least_function = "least"
-            elif isinstance(self.database_engine, Sqlite3Engine):
-                least_function = "min"
-            else:
-                raise RuntimeError("Unknown database engine")
-
             txn.execute(
-                sql % (least_function,),
+                sql,
                 (
                     room_id,
+                    current_depth,
                     self._clock.time_msec(),
-                    1000 * BACKFILL_EVENT_EXPONENTIAL_BACKOFF_STEP_SECONDS,
-                    1000 * BACKFILL_EVENT_BACKOFF_UPPER_BOUND_SECONDS,
+                    BACKFILL_EVENT_EXPONENTIAL_BACKOFF_MAXIMUM_DOUBLING_STEPS,
+                    BACKFILL_EVENT_EXPONENTIAL_BACKOFF_STEP_MILLISECONDS,
+                    limit,
                 ),
             )
             return cast(List[Tuple[str, int]], txn.fetchall())
diff --git a/synapse/storage/databases/main/event_push_actions.py b/synapse/storage/databases/main/event_push_actions.py
index f4cdc2e399..3fdf128d9e 100644
--- a/synapse/storage/databases/main/event_push_actions.py
+++ b/synapse/storage/databases/main/event_push_actions.py
@@ -366,14 +366,11 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas
         user_id: str,
     ) -> NotifCounts:
         # Get the stream ordering of the user's latest receipt in the room.
-        result = self.get_last_receipt_for_user_txn(
+        result = self.get_last_unthreaded_receipt_for_user_txn(
             txn,
             user_id,
             room_id,
-            receipt_types=(
-                ReceiptTypes.READ,
-                ReceiptTypes.READ_PRIVATE,
-            ),
+            receipt_types=(ReceiptTypes.READ, ReceiptTypes.READ_PRIVATE),
         )
 
         if result:
@@ -574,10 +571,7 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas
         receipt_types_clause, args = make_in_list_sql_clause(
             self.database_engine,
             "receipt_type",
-            (
-                ReceiptTypes.READ,
-                ReceiptTypes.READ_PRIVATE,
-            ),
+            (ReceiptTypes.READ, ReceiptTypes.READ_PRIVATE),
         )
 
         sql = f"""
@@ -1074,7 +1068,7 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas
                 limit,
             ),
         )
-        rows = txn.fetchall()
+        rows = cast(List[Tuple[int, str, str, int]], txn.fetchall())
 
         # For each new read receipt we delete push actions from before it and
         # recalculate the summary.
@@ -1119,18 +1113,18 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas
         # We always update `event_push_summary_last_receipt_stream_id` to
         # ensure that we don't rescan the same receipts for remote users.
 
-        upper_limit = max_receipts_stream_id
+        receipts_last_processed_stream_id = max_receipts_stream_id
         if len(rows) >= limit:
             # If we pulled out a limited number of rows we only update the
             # position to the last receipt we processed, so we continue
             # processing the rest next iteration.
-            upper_limit = rows[-1][0]
+            receipts_last_processed_stream_id = rows[-1][0]
 
         self.db_pool.simple_update_txn(
             txn,
             table="event_push_summary_last_receipt_stream_id",
             keyvalues={},
-            updatevalues={"stream_id": upper_limit},
+            updatevalues={"stream_id": receipts_last_processed_stream_id},
         )
 
         return len(rows) < limit
diff --git a/synapse/storage/databases/main/events.py b/synapse/storage/databases/main/events.py
index 0b86a53085..f707f74b50 100644
--- a/synapse/storage/databases/main/events.py
+++ b/synapse/storage/databases/main/events.py
@@ -1202,6 +1202,12 @@ class PersistEventsStore:
                 txn, room_id, members_changed
             )
 
+            # Check if any of the remote membership changes requires us to
+            # unsubscribe from their device lists.
+            self.store.handle_potentially_left_users_txn(
+                txn, {m for m in members_changed if not self.hs.is_mine_id(m)}
+            )
+
     def _upsert_room_version_txn(self, txn: LoggingTransaction, room_id: str) -> None:
         """Update the room version in the database based off current state
         events.
@@ -2128,13 +2134,13 @@ class PersistEventsStore:
                 appear in events_and_context.
         """
 
-        # Only non outlier events will have push actions associated with them,
+        # Only notifiable events will have push actions associated with them,
         # so let's filter them out. (This makes joining large rooms faster, as
         # these queries took seconds to process all the state events).
-        non_outlier_events = [
+        notifiable_events = [
             event
             for event, _ in events_and_contexts
-            if not event.internal_metadata.is_outlier()
+            if event.internal_metadata.is_notifiable()
         ]
 
         sql = """
@@ -2147,7 +2153,7 @@ class PersistEventsStore:
             WHERE event_id = ?
         """
 
-        if non_outlier_events:
+        if notifiable_events:
             txn.execute_batch(
                 sql,
                 (
@@ -2157,7 +2163,7 @@ class PersistEventsStore:
                         event.depth,
                         event.event_id,
                     )
-                    for event in non_outlier_events
+                    for event in notifiable_events
                 ),
             )
 
diff --git a/synapse/storage/databases/main/events_worker.py b/synapse/storage/databases/main/events_worker.py
index 9f6b1fcef1..74991a4992 100644
--- a/synapse/storage/databases/main/events_worker.py
+++ b/synapse/storage/databases/main/events_worker.py
@@ -1474,32 +1474,38 @@ class EventsWorkerStore(SQLBaseStore):
         # the batches as big as possible.
 
         results: Set[str] = set()
-        for chunk in batch_iter(event_ids, 500):
-            r = await self._have_seen_events_dict(
-                [(room_id, event_id) for event_id in chunk]
+        for event_ids_chunk in batch_iter(event_ids, 500):
+            events_seen_dict = await self._have_seen_events_dict(
+                room_id, event_ids_chunk
+            )
+            results.update(
+                eid for (eid, have_event) in events_seen_dict.items() if have_event
             )
-            results.update(eid for ((_rid, eid), have_event) in r.items() if have_event)
 
         return results
 
-    @cachedList(cached_method_name="have_seen_event", list_name="keys")
+    @cachedList(cached_method_name="have_seen_event", list_name="event_ids")
     async def _have_seen_events_dict(
-        self, keys: Collection[Tuple[str, str]]
-    ) -> Dict[Tuple[str, str], bool]:
+        self,
+        room_id: str,
+        event_ids: Collection[str],
+    ) -> Dict[str, bool]:
         """Helper for have_seen_events
 
         Returns:
-             a dict {(room_id, event_id)-> bool}
+             a dict {event_id -> bool}
         """
         # if the event cache contains the event, obviously we've seen it.
 
         cache_results = {
-            (rid, eid)
-            for (rid, eid) in keys
-            if await self._get_event_cache.contains((eid,))
+            event_id
+            for event_id in event_ids
+            if await self._get_event_cache.contains((event_id,))
         }
         results = dict.fromkeys(cache_results, True)
-        remaining = [k for k in keys if k not in cache_results]
+        remaining = [
+            event_id for event_id in event_ids if event_id not in cache_results
+        ]
         if not remaining:
             return results
 
@@ -1511,23 +1517,21 @@ class EventsWorkerStore(SQLBaseStore):
 
             sql = "SELECT event_id FROM events AS e WHERE "
             clause, args = make_in_list_sql_clause(
-                txn.database_engine, "e.event_id", [eid for (_rid, eid) in remaining]
+                txn.database_engine, "e.event_id", remaining
             )
             txn.execute(sql + clause, args)
             found_events = {eid for eid, in txn}
 
             # ... and then we can update the results for each key
-            results.update(
-                {(rid, eid): (eid in found_events) for (rid, eid) in remaining}
-            )
+            results.update({eid: (eid in found_events) for eid in remaining})
 
         await self.db_pool.runInteraction("have_seen_events", have_seen_events_txn)
         return results
 
     @cached(max_entries=100000, tree=True)
     async def have_seen_event(self, room_id: str, event_id: str) -> bool:
-        res = await self._have_seen_events_dict(((room_id, event_id),))
-        return res[(room_id, event_id)]
+        res = await self._have_seen_events_dict(room_id, [event_id])
+        return res[event_id]
 
     def _get_current_state_event_counts_txn(
         self, txn: LoggingTransaction, room_id: str
diff --git a/synapse/storage/databases/main/receipts.py b/synapse/storage/databases/main/receipts.py
index 52fe0db924..246f78ac1f 100644
--- a/synapse/storage/databases/main/receipts.py
+++ b/synapse/storage/databases/main/receipts.py
@@ -135,34 +135,7 @@ class ReceiptsWorkerStore(SQLBaseStore):
         """Get the current max stream ID for receipts stream"""
         return self._receipts_id_gen.get_current_token()
 
-    async def get_last_receipt_event_id_for_user(
-        self, user_id: str, room_id: str, receipt_types: Collection[str]
-    ) -> Optional[str]:
-        """
-        Fetch the event ID for the latest receipt in a room with one of the given receipt types.
-
-        Args:
-            user_id: The user to fetch receipts for.
-            room_id: The room ID to fetch the receipt for.
-            receipt_type: The receipt types to fetch.
-
-        Returns:
-            The latest receipt, if one exists.
-        """
-        result = await self.db_pool.runInteraction(
-            "get_last_receipt_event_id_for_user",
-            self.get_last_receipt_for_user_txn,
-            user_id,
-            room_id,
-            receipt_types,
-        )
-        if not result:
-            return None
-
-        event_id, _ = result
-        return event_id
-
-    def get_last_receipt_for_user_txn(
+    def get_last_unthreaded_receipt_for_user_txn(
         self,
         txn: LoggingTransaction,
         user_id: str,
@@ -170,13 +143,13 @@ class ReceiptsWorkerStore(SQLBaseStore):
         receipt_types: Collection[str],
     ) -> Optional[Tuple[str, int]]:
         """
-        Fetch the event ID and stream_ordering for the latest receipt in a room
-        with one of the given receipt types.
+        Fetch the event ID and stream_ordering for the latest unthreaded receipt
+        in a room with one of the given receipt types.
 
         Args:
             user_id: The user to fetch receipts for.
             room_id: The room ID to fetch the receipt for.
-            receipt_type: The receipt types to fetch.
+            receipt_types: The receipt types to fetch.
 
         Returns:
             The event ID and stream ordering of the latest receipt, if one exists.
@@ -193,6 +166,7 @@ class ReceiptsWorkerStore(SQLBaseStore):
             WHERE {clause}
             AND user_id = ?
             AND room_id = ?
+            AND thread_id IS NULL
             ORDER BY stream_ordering DESC
             LIMIT 1
         """
diff --git a/synapse/storage/databases/main/registration.py b/synapse/storage/databases/main/registration.py
index ac821878b0..2996d6bb4d 100644
--- a/synapse/storage/databases/main/registration.py
+++ b/synapse/storage/databases/main/registration.py
@@ -166,27 +166,49 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore):
     @cached()
     async def get_user_by_id(self, user_id: str) -> Optional[Dict[str, Any]]:
         """Deprecated: use get_userinfo_by_id instead"""
-        return await self.db_pool.simple_select_one(
-            table="users",
-            keyvalues={"name": user_id},
-            retcols=[
-                "name",
-                "password_hash",
-                "is_guest",
-                "admin",
-                "consent_version",
-                "consent_ts",
-                "consent_server_notice_sent",
-                "appservice_id",
-                "creation_ts",
-                "user_type",
-                "deactivated",
-                "shadow_banned",
-            ],
-            allow_none=True,
+
+        def get_user_by_id_txn(txn: LoggingTransaction) -> Optional[Dict[str, Any]]:
+            # We could technically use simple_select_one here, but it would not perform
+            # the COALESCEs (unless hacked into the column names), which could yield
+            # confusing results.
+            txn.execute(
+                """
+                SELECT
+                    name, password_hash, is_guest, admin, consent_version, consent_ts,
+                    consent_server_notice_sent, appservice_id, creation_ts, user_type,
+                    deactivated, COALESCE(shadow_banned, FALSE) AS shadow_banned,
+                    COALESCE(approved, TRUE) AS approved
+                FROM users
+                WHERE name = ?
+                """,
+                (user_id,),
+            )
+
+            rows = self.db_pool.cursor_to_dict(txn)
+
+            if len(rows) == 0:
+                return None
+
+            return rows[0]
+
+        row = await self.db_pool.runInteraction(
             desc="get_user_by_id",
+            func=get_user_by_id_txn,
         )
 
+        if row is not None:
+            # If we're using SQLite our boolean values will be integers. Because we
+            # present some of this data as is to e.g. server admins via REST APIs, we
+            # want to make sure we're returning the right type of data.
+            # Note: when adding a column name to this list, be wary of NULLable columns,
+            # since NULL values will be turned into False.
+            boolean_columns = ["admin", "deactivated", "shadow_banned", "approved"]
+            for column in boolean_columns:
+                if not isinstance(row[column], bool):
+                    row[column] = bool(row[column])
+
+        return row
+
     async def get_userinfo_by_id(self, user_id: str) -> Optional[UserInfo]:
         """Get a UserInfo object for a user by user ID.
 
@@ -1779,6 +1801,40 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore):
 
         return res if res else False
 
+    @cached()
+    async def is_user_approved(self, user_id: str) -> bool:
+        """Checks if a user is approved and therefore can be allowed to log in.
+
+        If the user's 'approved' column is NULL, we consider it as true given it means
+        the user was registered when support for an approval flow was either disabled
+        or nonexistent.
+
+        Args:
+            user_id: the user to check the approval status of.
+
+        Returns:
+            A boolean that is True if the user is approved, False otherwise.
+        """
+
+        def is_user_approved_txn(txn: LoggingTransaction) -> bool:
+            txn.execute(
+                """
+                SELECT COALESCE(approved, TRUE) AS approved FROM users WHERE name = ?
+                """,
+                (user_id,),
+            )
+
+            rows = self.db_pool.cursor_to_dict(txn)
+
+            # We cast to bool because the value returned by the database engine might
+            # be an integer if we're using SQLite.
+            return bool(rows[0]["approved"])
+
+        return await self.db_pool.runInteraction(
+            desc="is_user_pending_approval",
+            func=is_user_approved_txn,
+        )
+
 
 class RegistrationBackgroundUpdateStore(RegistrationWorkerStore):
     def __init__(
@@ -1916,6 +1972,29 @@ class RegistrationBackgroundUpdateStore(RegistrationWorkerStore):
         self._invalidate_cache_and_stream(txn, self.get_user_by_id, (user_id,))
         txn.call_after(self.is_guest.invalidate, (user_id,))
 
+    def update_user_approval_status_txn(
+        self, txn: LoggingTransaction, user_id: str, approved: bool
+    ) -> None:
+        """Set the user's 'approved' flag to the given value.
+
+        The boolean is turned into an int because the column is a smallint.
+
+        Args:
+            txn: the current database transaction.
+            user_id: the user to update the flag for.
+            approved: the value to set the flag to.
+        """
+        self.db_pool.simple_update_one_txn(
+            txn=txn,
+            table="users",
+            keyvalues={"name": user_id},
+            updatevalues={"approved": approved},
+        )
+
+        # Invalidate the caches of methods that read the value of the 'approved' flag.
+        self._invalidate_cache_and_stream(txn, self.get_user_by_id, (user_id,))
+        self._invalidate_cache_and_stream(txn, self.is_user_approved, (user_id,))
+
 
 class RegistrationStore(StatsStore, RegistrationBackgroundUpdateStore):
     def __init__(
@@ -1933,6 +2012,13 @@ class RegistrationStore(StatsStore, RegistrationBackgroundUpdateStore):
         self._access_tokens_id_gen = IdGenerator(db_conn, "access_tokens", "id")
         self._refresh_tokens_id_gen = IdGenerator(db_conn, "refresh_tokens", "id")
 
+        # If support for MSC3866 is enabled and configured to require approval for new
+        # account, we will create new users with an 'approved' flag set to false.
+        self._require_approval = (
+            hs.config.experimental.msc3866.enabled
+            and hs.config.experimental.msc3866.require_approval_for_new_accounts
+        )
+
     async def add_access_token_to_user(
         self,
         user_id: str,
@@ -2065,6 +2151,7 @@ class RegistrationStore(StatsStore, RegistrationBackgroundUpdateStore):
         admin: bool = False,
         user_type: Optional[str] = None,
         shadow_banned: bool = False,
+        approved: bool = False,
     ) -> None:
         """Attempts to register an account.
 
@@ -2083,6 +2170,8 @@ class RegistrationStore(StatsStore, RegistrationBackgroundUpdateStore):
                 or None for a normal user.
             shadow_banned: Whether the user is shadow-banned, i.e. they may be
                 told their requests succeeded but we ignore them.
+            approved: Whether to consider the user has already been approved by an
+                administrator.
 
         Raises:
             StoreError if the user_id could not be registered.
@@ -2099,6 +2188,7 @@ class RegistrationStore(StatsStore, RegistrationBackgroundUpdateStore):
             admin,
             user_type,
             shadow_banned,
+            approved,
         )
 
     def _register_user(
@@ -2113,11 +2203,14 @@ class RegistrationStore(StatsStore, RegistrationBackgroundUpdateStore):
         admin: bool,
         user_type: Optional[str],
         shadow_banned: bool,
+        approved: bool,
     ) -> None:
         user_id_obj = UserID.from_string(user_id)
 
         now = int(self._clock.time())
 
+        user_approved = approved or not self._require_approval
+
         try:
             if was_guest:
                 # Ensure that the guest user actually exists
@@ -2143,6 +2236,7 @@ class RegistrationStore(StatsStore, RegistrationBackgroundUpdateStore):
                         "admin": 1 if admin else 0,
                         "user_type": user_type,
                         "shadow_banned": shadow_banned,
+                        "approved": user_approved,
                     },
                 )
             else:
@@ -2158,6 +2252,7 @@ class RegistrationStore(StatsStore, RegistrationBackgroundUpdateStore):
                         "admin": 1 if admin else 0,
                         "user_type": user_type,
                         "shadow_banned": shadow_banned,
+                        "approved": user_approved,
                     },
                 )
 
@@ -2503,6 +2598,25 @@ class RegistrationStore(StatsStore, RegistrationBackgroundUpdateStore):
             start_or_continue_validation_session_txn,
         )
 
+    async def update_user_approval_status(
+        self, user_id: UserID, approved: bool
+    ) -> None:
+        """Set the user's 'approved' flag to the given value.
+
+        The boolean will be turned into an int (in update_user_approval_status_txn)
+        because the column is a smallint.
+
+        Args:
+            user_id: the user to update the flag for.
+            approved: the value to set the flag to.
+        """
+        await self.db_pool.runInteraction(
+            "update_user_approval_status",
+            self.update_user_approval_status_txn,
+            user_id.to_string(),
+            approved,
+        )
+
 
 def find_max_generated_user_id_localpart(cur: Cursor) -> int:
     """
diff --git a/synapse/storage/databases/main/room.py b/synapse/storage/databases/main/room.py
index 5dd116d766..059eef5c22 100644
--- a/synapse/storage/databases/main/room.py
+++ b/synapse/storage/databases/main/room.py
@@ -1217,6 +1217,26 @@ class RoomWorkerStore(CacheInvalidationWorkerStore):
         )
         self._invalidate_cache_and_stream(txn, self.is_partial_state_room, (room_id,))
 
+        # We now delete anything from `device_lists_remote_pending` with a
+        # stream ID less than the minimum
+        # `partial_state_rooms.device_lists_stream_id`, as we no longer need them.
+        device_lists_stream_id = DatabasePool.simple_select_one_onecol_txn(
+            txn,
+            table="partial_state_rooms",
+            keyvalues={},
+            retcol="MIN(device_lists_stream_id)",
+            allow_none=True,
+        )
+        if device_lists_stream_id is None:
+            # There are no rooms being currently partially joined, so we delete everything.
+            txn.execute("DELETE FROM device_lists_remote_pending")
+        else:
+            sql = """
+                DELETE FROM device_lists_remote_pending
+                WHERE stream_id <= ?
+            """
+            txn.execute(sql, (device_lists_stream_id,))
+
     @cached()
     async def is_partial_state_room(self, room_id: str) -> bool:
         """Checks if this room has partial state.
@@ -1236,6 +1256,22 @@ class RoomWorkerStore(CacheInvalidationWorkerStore):
 
         return entry is not None
 
+    async def get_join_event_id_and_device_lists_stream_id_for_partial_state(
+        self, room_id: str
+    ) -> Tuple[str, int]:
+        """Get the event ID of the initial join that started the partial
+        join, and the device list stream ID at the point we started the partial
+        join.
+        """
+
+        result = await self.db_pool.simple_select_one(
+            table="partial_state_rooms",
+            keyvalues={"room_id": room_id},
+            retcols=("join_event_id", "device_lists_stream_id"),
+            desc="get_join_event_id_for_partial_state",
+        )
+        return result["join_event_id"], result["device_lists_stream_id"]
+
 
 class _BackgroundUpdates:
     REMOVE_TOMESTONED_ROOMS_BG_UPDATE = "remove_tombstoned_rooms_from_directory"
@@ -1777,28 +1813,46 @@ class RoomStore(RoomBackgroundUpdateStore, RoomWorkerStore):
         self,
         room_id: str,
         servers: Collection[str],
+        device_lists_stream_id: int,
     ) -> None:
-        """Mark the given room as containing events with partial state
+        """Mark the given room as containing events with partial state.
+
+        We also store additional data that describes _when_ we first partial-joined this
+        room, which helps us to keep other homeservers in sync when we finally fully
+        join this room.
+
+        We do not include a `join_event_id` here---we need to wait for the join event
+        to be persisted first.
 
         Args:
             room_id: the ID of the room
             servers: other servers known to be in the room
+            device_lists_stream_id: the device_lists stream ID at the time when we first
+                joined the room.
         """
         await self.db_pool.runInteraction(
             "store_partial_state_room",
             self._store_partial_state_room_txn,
             room_id,
             servers,
+            device_lists_stream_id,
         )
 
     def _store_partial_state_room_txn(
-        self, txn: LoggingTransaction, room_id: str, servers: Collection[str]
+        self,
+        txn: LoggingTransaction,
+        room_id: str,
+        servers: Collection[str],
+        device_lists_stream_id: int,
     ) -> None:
         DatabasePool.simple_insert_txn(
             txn,
             table="partial_state_rooms",
             values={
                 "room_id": room_id,
+                "device_lists_stream_id": device_lists_stream_id,
+                # To be updated later once the join event is persisted.
+                "join_event_id": None,
             },
         )
         DatabasePool.simple_insert_many_txn(
@@ -1809,6 +1863,36 @@ class RoomStore(RoomBackgroundUpdateStore, RoomWorkerStore):
         )
         self._invalidate_cache_and_stream(txn, self.is_partial_state_room, (room_id,))
 
+    async def write_partial_state_rooms_join_event_id(
+        self,
+        room_id: str,
+        join_event_id: str,
+    ) -> None:
+        """Record the join event which resulted from a partial join.
+
+        We do this separately to `store_partial_state_room` because we need to wait for
+        the join event to be persisted. Otherwise we violate a foreign key constraint.
+        """
+        await self.db_pool.runInteraction(
+            "write_partial_state_rooms_join_event_id",
+            self._write_partial_state_rooms_join_event_id,
+            room_id,
+            join_event_id,
+        )
+
+    def _write_partial_state_rooms_join_event_id(
+        self,
+        txn: LoggingTransaction,
+        room_id: str,
+        join_event_id: str,
+    ) -> None:
+        DatabasePool.simple_update_txn(
+            txn,
+            table="partial_state_rooms",
+            keyvalues={"room_id": room_id},
+            updatevalues={"join_event_id": join_event_id},
+        )
+
     async def maybe_store_room_on_outlier_membership(
         self, room_id: str, room_version: RoomVersion
     ) -> None:
diff --git a/synapse/storage/databases/main/roommember.py b/synapse/storage/databases/main/roommember.py
index a8d224602a..2337289d88 100644
--- a/synapse/storage/databases/main/roommember.py
+++ b/synapse/storage/databases/main/roommember.py
@@ -15,7 +15,6 @@
 import logging
 from typing import (
     TYPE_CHECKING,
-    Callable,
     Collection,
     Dict,
     FrozenSet,
@@ -52,7 +51,6 @@ from synapse.types import JsonDict, PersistedEventPosition, StateMap, get_domain
 from synapse.util.async_helpers import Linearizer
 from synapse.util.caches import intern_string
 from synapse.util.caches.descriptors import _CacheContext, cached, cachedList
-from synapse.util.cancellation import cancellable
 from synapse.util.iterutils import batch_iter
 from synapse.util.metrics import Measure
 
@@ -148,42 +146,37 @@ class RoomMemberWorkerStore(EventsWorkerStore):
 
     @cached(max_entries=100000, iterable=True)
     async def get_users_in_room(self, room_id: str) -> List[str]:
-        """
-        Returns a list of users in the room sorted by longest in the room first
-        (aka. with the lowest depth). This is done to match the sort in
-        `get_current_hosts_in_room()` and so we can re-use the cache but it's
-        not horrible to have here either.
-
-        Uses `m.room.member`s in the room state at the current forward extremities to
-        determine which users are in the room.
+        """Returns a list of users in the room.
 
         Will return inaccurate results for rooms with partial state, since the state for
         the forward extremities of those rooms will exclude most members. We may also
         calculate room state incorrectly for such rooms and believe that a member is or
         is not in the room when the opposite is true.
         """
-        return await self.db_pool.runInteraction(
-            "get_users_in_room", self.get_users_in_room_txn, room_id
+        return await self.db_pool.simple_select_onecol(
+            table="current_state_events",
+            keyvalues={
+                "type": EventTypes.Member,
+                "room_id": room_id,
+                "membership": Membership.JOIN,
+            },
+            retcol="state_key",
+            desc="get_users_in_room",
         )
 
     def get_users_in_room_txn(self, txn: LoggingTransaction, room_id: str) -> List[str]:
-        """
-        Returns a list of users in the room sorted by longest in the room first
-        (aka. with the lowest depth). This is done to match the sort in
-        `get_current_hosts_in_room()` and so we can re-use the cache but it's
-        not horrible to have here either.
-        """
-        sql = """
-            SELECT c.state_key FROM current_state_events as c
-            /* Get the depth of the event from the events table */
-            INNER JOIN events AS e USING (event_id)
-            WHERE c.type = 'm.room.member' AND c.room_id = ? AND membership = ?
-            /* Sorted by lowest depth first */
-            ORDER BY e.depth ASC;
-        """
+        """Returns a list of users in the room."""
 
-        txn.execute(sql, (room_id, Membership.JOIN))
-        return [r[0] for r in txn]
+        return self.db_pool.simple_select_onecol_txn(
+            txn,
+            table="current_state_events",
+            keyvalues={
+                "type": EventTypes.Member,
+                "room_id": room_id,
+                "membership": Membership.JOIN,
+            },
+            retcol="state_key",
+        )
 
     @cached()
     def get_user_in_room_with_profile(
@@ -600,106 +593,109 @@ class RoomMemberWorkerStore(EventsWorkerStore):
             for room_id, instance, stream_id in txn
         )
 
-    @cachedList(
-        cached_method_name="get_rooms_for_user_with_stream_ordering",
-        list_name="user_ids",
-    )
-    async def get_rooms_for_users_with_stream_ordering(
+    async def get_users_server_still_shares_room_with(
         self, user_ids: Collection[str]
-    ) -> Dict[str, FrozenSet[GetRoomsForUserWithStreamOrdering]]:
-        """A batched version of `get_rooms_for_user_with_stream_ordering`.
-
-        Returns:
-            Map from user_id to set of rooms that is currently in.
+    ) -> Set[str]:
+        """Given a list of users return the set that the server still share a
+        room with.
         """
-        return await self.db_pool.runInteraction(
-            "get_rooms_for_users_with_stream_ordering",
-            self._get_rooms_for_users_with_stream_ordering_txn,
-            user_ids,
-        )
 
-    def _get_rooms_for_users_with_stream_ordering_txn(
-        self, txn: LoggingTransaction, user_ids: Collection[str]
-    ) -> Dict[str, FrozenSet[GetRoomsForUserWithStreamOrdering]]:
+        if not user_ids:
+            return set()
 
-        clause, args = make_in_list_sql_clause(
-            self.database_engine,
-            "c.state_key",
+        return await self.db_pool.runInteraction(
+            "get_users_server_still_shares_room_with",
+            self.get_users_server_still_shares_room_with_txn,
             user_ids,
         )
 
-        sql = f"""
-            SELECT c.state_key, room_id, e.instance_name, e.stream_ordering
-            FROM current_state_events AS c
-            INNER JOIN events AS e USING (room_id, event_id)
+    def get_users_server_still_shares_room_with_txn(
+        self,
+        txn: LoggingTransaction,
+        user_ids: Collection[str],
+    ) -> Set[str]:
+        if not user_ids:
+            return set()
+
+        sql = """
+            SELECT state_key FROM current_state_events
             WHERE
-                c.type = 'm.room.member'
-                AND c.membership = ?
-                AND {clause}
+                type = 'm.room.member'
+                AND membership = 'join'
+                AND %s
+            GROUP BY state_key
         """
 
-        txn.execute(sql, [Membership.JOIN] + args)
+        clause, args = make_in_list_sql_clause(
+            self.database_engine, "state_key", user_ids
+        )
 
-        result: Dict[str, Set[GetRoomsForUserWithStreamOrdering]] = {
-            user_id: set() for user_id in user_ids
-        }
-        for user_id, room_id, instance, stream_id in txn:
-            result[user_id].add(
-                GetRoomsForUserWithStreamOrdering(
-                    room_id, PersistedEventPosition(instance, stream_id)
-                )
-            )
+        txn.execute(sql % (clause,), args)
 
-        return {user_id: frozenset(v) for user_id, v in result.items()}
+        return {row[0] for row in txn}
 
-    async def get_users_server_still_shares_room_with(
-        self, user_ids: Collection[str]
-    ) -> Set[str]:
-        """Given a list of users return the set that the server still share a
-        room with.
-        """
+    @cached(max_entries=500000, iterable=True)
+    async def get_rooms_for_user(self, user_id: str) -> FrozenSet[str]:
+        """Returns a set of room_ids the user is currently joined to.
 
-        if not user_ids:
-            return set()
+        If a remote user only returns rooms this server is currently
+        participating in.
+        """
+        rooms = self.get_rooms_for_user_with_stream_ordering.cache.get_immediate(
+            (user_id,),
+            None,
+            update_metrics=False,
+        )
+        if rooms:
+            return frozenset(r.room_id for r in rooms)
 
-        def _get_users_server_still_shares_room_with_txn(
-            txn: LoggingTransaction,
-        ) -> Set[str]:
-            sql = """
-                SELECT state_key FROM current_state_events
-                WHERE
-                    type = 'm.room.member'
-                    AND membership = 'join'
-                    AND %s
-                GROUP BY state_key
-            """
+        room_ids = await self.db_pool.simple_select_onecol(
+            table="current_state_events",
+            keyvalues={
+                "type": EventTypes.Member,
+                "membership": Membership.JOIN,
+                "state_key": user_id,
+            },
+            retcol="room_id",
+            desc="get_rooms_for_user",
+        )
 
-            clause, args = make_in_list_sql_clause(
-                self.database_engine, "state_key", user_ids
-            )
+        return frozenset(room_ids)
 
-            txn.execute(sql % (clause,), args)
+    @cachedList(
+        cached_method_name="get_rooms_for_user",
+        list_name="user_ids",
+    )
+    async def get_rooms_for_users(
+        self, user_ids: Collection[str]
+    ) -> Dict[str, FrozenSet[str]]:
+        """A batched version of `get_rooms_for_user`.
 
-            return {row[0] for row in txn}
+        Returns:
+            Map from user_id to set of rooms that is currently in.
+        """
 
-        return await self.db_pool.runInteraction(
-            "get_users_server_still_shares_room_with",
-            _get_users_server_still_shares_room_with_txn,
+        rows = await self.db_pool.simple_select_many_batch(
+            table="current_state_events",
+            column="state_key",
+            iterable=user_ids,
+            retcols=(
+                "state_key",
+                "room_id",
+            ),
+            keyvalues={
+                "type": EventTypes.Member,
+                "membership": Membership.JOIN,
+            },
+            desc="get_rooms_for_users",
         )
 
-    @cancellable
-    async def get_rooms_for_user(
-        self, user_id: str, on_invalidate: Optional[Callable[[], None]] = None
-    ) -> FrozenSet[str]:
-        """Returns a set of room_ids the user is currently joined to.
+        user_rooms: Dict[str, Set[str]] = {user_id: set() for user_id in user_ids}
 
-        If a remote user only returns rooms this server is currently
-        participating in.
-        """
-        rooms = await self.get_rooms_for_user_with_stream_ordering(
-            user_id, on_invalidate=on_invalidate
-        )
-        return frozenset(r.room_id for r in rooms)
+        for row in rows:
+            user_rooms[row["state_key"]].add(row["room_id"])
+
+        return {key: frozenset(rooms) for key, rooms in user_rooms.items()}
 
     @cached(max_entries=10000)
     async def does_pair_of_users_share_a_room(
@@ -930,7 +926,44 @@ class RoomMemberWorkerStore(EventsWorkerStore):
         return True
 
     @cached(iterable=True, max_entries=10000)
-    async def get_current_hosts_in_room(self, room_id: str) -> List[str]:
+    async def get_current_hosts_in_room(self, room_id: str) -> Set[str]:
+        """Get current hosts in room based on current state."""
+
+        # First we check if we already have `get_users_in_room` in the cache, as
+        # we can just calculate result from that
+        users = self.get_users_in_room.cache.get_immediate(
+            (room_id,), None, update_metrics=False
+        )
+        if users is not None:
+            return {get_domain_from_id(u) for u in users}
+
+        if isinstance(self.database_engine, Sqlite3Engine):
+            # If we're using SQLite then let's just always use
+            # `get_users_in_room` rather than funky SQL.
+            users = await self.get_users_in_room(room_id)
+            return {get_domain_from_id(u) for u in users}
+
+        # For PostgreSQL we can use a regex to pull out the domains from the
+        # joined users in `current_state_events` via regex.
+
+        def get_current_hosts_in_room_txn(txn: LoggingTransaction) -> Set[str]:
+            sql = """
+                SELECT DISTINCT substring(state_key FROM '@[^:]*:(.*)$')
+                FROM current_state_events
+                WHERE
+                    type = 'm.room.member'
+                    AND membership = 'join'
+                    AND room_id = ?
+            """
+            txn.execute(sql, (room_id,))
+            return {d for d, in txn}
+
+        return await self.db_pool.runInteraction(
+            "get_current_hosts_in_room", get_current_hosts_in_room_txn
+        )
+
+    @cached(iterable=True, max_entries=10000)
+    async def get_current_hosts_in_room_ordered(self, room_id: str) -> List[str]:
         """
         Get current hosts in room based on current state.
 
@@ -938,48 +971,33 @@ class RoomMemberWorkerStore(EventsWorkerStore):
         longest is good because they're most likely to have anything we ask
         about.
 
-        Uses `m.room.member`s in the room state at the current forward extremities to
-        determine which hosts are in the room.
+        For SQLite the returned list is not ordered, as SQLite doesn't support
+        the appropriate SQL.
 
-        Will return inaccurate results for rooms with partial state, since the state for
-        the forward extremities of those rooms will exclude most members. We may also
-        calculate room state incorrectly for such rooms and believe that a host is or
-        is not in the room when the opposite is true.
+        Uses `m.room.member`s in the room state at the current forward
+        extremities to determine which hosts are in the room.
+
+        Will return inaccurate results for rooms with partial state, since the
+        state for the forward extremities of those rooms will exclude most
+        members. We may also calculate room state incorrectly for such rooms and
+        believe that a host is or is not in the room when the opposite is true.
 
         Returns:
             Returns a list of servers sorted by longest in the room first. (aka.
             sorted by join with the lowest depth first).
         """
 
-        # First we check if we already have `get_users_in_room` in the cache, as
-        # we can just calculate result from that
-        users = self.get_users_in_room.cache.get_immediate(
-            (room_id,), None, update_metrics=False
-        )
-        if users is None and isinstance(self.database_engine, Sqlite3Engine):
+        if isinstance(self.database_engine, Sqlite3Engine):
             # If we're using SQLite then let's just always use
             # `get_users_in_room` rather than funky SQL.
-            users = await self.get_users_in_room(room_id)
 
-        if users is not None:
-            # Because `users` is sorted from lowest -> highest depth, the list
-            # of domains will also be sorted that way.
-            domains: List[str] = []
-            # We use a `Set` just for fast lookups
-            domain_set: Set[str] = set()
-            for u in users:
-                if ":" not in u:
-                    continue
-                domain = get_domain_from_id(u)
-                if domain not in domain_set:
-                    domain_set.add(domain)
-                    domains.append(domain)
-            return domains
+            domains = await self.get_current_hosts_in_room(room_id)
+            return list(domains)
 
         # For PostgreSQL we can use a regex to pull out the domains from the
         # joined users in `current_state_events` via regex.
 
-        def get_current_hosts_in_room_txn(txn: LoggingTransaction) -> List[str]:
+        def get_current_hosts_in_room_ordered_txn(txn: LoggingTransaction) -> List[str]:
             # Returns a list of servers currently joined in the room sorted by
             # longest in the room first (aka. with the lowest depth). The
             # heuristic of sorting by servers who have been in the room the
@@ -1007,7 +1025,7 @@ class RoomMemberWorkerStore(EventsWorkerStore):
             return [d for d, in txn if d is not None]
 
         return await self.db_pool.runInteraction(
-            "get_current_hosts_in_room", get_current_hosts_in_room_txn
+            "get_current_hosts_in_room_ordered", get_current_hosts_in_room_ordered_txn
         )
 
     async def get_joined_hosts(
diff --git a/synapse/storage/databases/main/search.py b/synapse/storage/databases/main/search.py
index f6e24b68d2..1b79acf955 100644
--- a/synapse/storage/databases/main/search.py
+++ b/synapse/storage/databases/main/search.py
@@ -641,7 +641,7 @@ class SearchStore(SearchBackgroundUpdateStore):
             raise Exception("Unrecognized database engine")
 
         # mypy expects to append only a `str`, not an `int`
-        args.append(limit)  # type: ignore[arg-type]
+        args.append(limit)
 
         results = await self.db_pool.execute(
             "search_rooms", self.db_pool.cursor_to_dict, sql, *args
diff --git a/synapse/storage/schema/main/delta/73/03users_approved_column.sql b/synapse/storage/schema/main/delta/73/03users_approved_column.sql
new file mode 100644
index 0000000000..5328d592ea
--- /dev/null
+++ b/synapse/storage/schema/main/delta/73/03users_approved_column.sql
@@ -0,0 +1,20 @@
+/* Copyright 2022 The Matrix.org Foundation C.I.C
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *    http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+-- Add a column to the users table to track whether the user needs to be approved by an
+-- administrator.
+-- A NULL column means the user was created before this feature was supported by Synapse,
+-- and should be considered as TRUE.
+ALTER TABLE users ADD COLUMN approved BOOLEAN;
diff --git a/synapse/storage/schema/main/delta/73/04partial_join_details.sql b/synapse/storage/schema/main/delta/73/04partial_join_details.sql
new file mode 100644
index 0000000000..5fb2bfe1a2
--- /dev/null
+++ b/synapse/storage/schema/main/delta/73/04partial_join_details.sql
@@ -0,0 +1,23 @@
+/* Copyright 2022 The Matrix.org Foundation C.I.C
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *    http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+-- To ensure we correctly notify other homeservers about device list changes from our
+-- users after a partial join transitions to a full join, we need to know when we began
+-- the partial join. For now it's sufficient to know the device_list stream_id at the
+-- time of the partial join, and the join event created for us during a partial join.
+--
+-- Both columns are backwards compatible.
+ALTER TABLE partial_state_rooms ADD COLUMN device_lists_stream_id BIGINT NOT NULL DEFAULT 0;
+ALTER TABLE partial_state_rooms ADD COLUMN join_event_id TEXT REFERENCES events(event_id);
diff --git a/synapse/storage/schema/main/delta/73/04pending_device_list_updates.sql b/synapse/storage/schema/main/delta/73/04pending_device_list_updates.sql
new file mode 100644
index 0000000000..dbd78d677d
--- /dev/null
+++ b/synapse/storage/schema/main/delta/73/04pending_device_list_updates.sql
@@ -0,0 +1,28 @@
+/* Copyright 2022 The Matrix.org Foundation C.I.C
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *    http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+-- Stores remote device lists we have received for remote users while a partial
+-- join is in progress.
+--
+-- This allows us to replay any device list updates if it turns out the remote
+-- user was in the partially joined room
+CREATE TABLE device_lists_remote_pending(
+    stream_id BIGINT PRIMARY KEY,
+    user_id TEXT NOT NULL,
+    device_id TEXT NOT NULL
+);
+
+-- We only keep the most recent update for a given user/device pair.
+CREATE UNIQUE INDEX device_lists_remote_pending_user_device_id ON device_lists_remote_pending(user_id, device_id);
diff --git a/synapse/util/caches/descriptors.py b/synapse/util/caches/descriptors.py
index 3909f1caea..0391966462 100644
--- a/synapse/util/caches/descriptors.py
+++ b/synapse/util/caches/descriptors.py
@@ -431,6 +431,12 @@ class DeferredCacheListDescriptor(_CacheDescriptorBase):
         cache: DeferredCache[CacheKey, Any] = cached_method.cache
         num_args = cached_method.num_args
 
+        if num_args != self.num_args:
+            raise Exception(
+                "Number of args (%s) does not match underlying cache_method_name=%s (%s)."
+                % (self.num_args, self.cached_method_name, num_args)
+            )
+
         @functools.wraps(self.orig)
         def wrapped(*args: Any, **kwargs: Any) -> "defer.Deferred[Dict]":
             # If we're passed a cache_context then we'll want to call its
diff --git a/synapse/util/check_dependencies.py b/synapse/util/check_dependencies.py
index 66f1da7502..3b1e205700 100644
--- a/synapse/util/check_dependencies.py
+++ b/synapse/util/check_dependencies.py
@@ -66,6 +66,21 @@ def _is_dev_dependency(req: Requirement) -> bool:
     )
 
 
+def _should_ignore_runtime_requirement(req: Requirement) -> bool:
+    # This is a build-time dependency. Irritatingly, `poetry build` ignores the
+    # requirements listed in the [build-system] section of pyproject.toml, so in order
+    # to support `poetry install --no-dev` we have to mark it as a runtime dependency.
+    # See discussion on https://github.com/python-poetry/poetry/issues/6154 (it sounds
+    # like the poetry authors don't consider this a bug?)
+    #
+    # In any case, workaround this by ignoring setuptools_rust here. (It might be
+    # slightly cleaner to put `setuptools_rust` in a `build` extra or similar, but for
+    # now let's do something quick and dirty.
+    if req.name == "setuptools_rust":
+        return True
+    return False
+
+
 class Dependency(NamedTuple):
     requirement: Requirement
     must_be_installed: bool
@@ -77,7 +92,7 @@ def _generic_dependencies() -> Iterable[Dependency]:
     assert requirements is not None
     for raw_requirement in requirements:
         req = Requirement(raw_requirement)
-        if _is_dev_dependency(req):
+        if _is_dev_dependency(req) or _should_ignore_runtime_requirement(req):
             continue
 
         # https://packaging.pypa.io/en/latest/markers.html#usage notes that
diff --git a/synapse/visibility.py b/synapse/visibility.py
index 342d60a921..1d3a1a940f 100644
--- a/synapse/visibility.py
+++ b/synapse/visibility.py
@@ -162,6 +162,10 @@ async def filter_event_for_clients_with_state(
     if event.internal_metadata.is_soft_failed():
         return []
 
+    # Fast path if we don't have any user IDs to check.
+    if not user_ids:
+        return ()
+
     # Make a set for all user IDs that haven't been filtered out by a check.
     allowed_user_ids = set(user_ids)
 
diff --git a/tests/handlers/test_message.py b/tests/handlers/test_message.py
index 986b50ce0c..99384837d0 100644
--- a/tests/handlers/test_message.py
+++ b/tests/handlers/test_message.py
@@ -105,7 +105,10 @@ class EventCreationTestCase(unittest.HomeserverTestCase):
         event1, context = self._create_duplicate_event(txn_id)
 
         ret_event1 = self.get_success(
-            self.handler.handle_new_client_event(self.requester, event1, context)
+            self.handler.handle_new_client_event(
+                self.requester,
+                events_and_context=[(event1, context)],
+            )
         )
         stream_id1 = ret_event1.internal_metadata.stream_ordering
 
@@ -118,7 +121,10 @@ class EventCreationTestCase(unittest.HomeserverTestCase):
         self.assertNotEqual(event1.event_id, event2.event_id)
 
         ret_event2 = self.get_success(
-            self.handler.handle_new_client_event(self.requester, event2, context)
+            self.handler.handle_new_client_event(
+                self.requester,
+                events_and_context=[(event2, context)],
+            )
         )
         stream_id2 = ret_event2.internal_metadata.stream_ordering
 
diff --git a/tests/handlers/test_register.py b/tests/handlers/test_register.py
index 86b3d51975..765df75d91 100644
--- a/tests/handlers/test_register.py
+++ b/tests/handlers/test_register.py
@@ -497,7 +497,9 @@ class RegistrationTestCase(unittest.HomeserverTestCase):
             )
         )
         self.get_success(
-            event_creation_handler.handle_new_client_event(requester, event, context)
+            event_creation_handler.handle_new_client_event(
+                requester, events_and_context=[(event, context)]
+            )
         )
 
         # Register a second user, which won't be be in the room (or even have an invite)
diff --git a/tests/handlers/test_sync.py b/tests/handlers/test_sync.py
index e3f38fbcc5..ab5c101eb7 100644
--- a/tests/handlers/test_sync.py
+++ b/tests/handlers/test_sync.py
@@ -159,6 +159,7 @@ class SyncTestCase(tests.unittest.HomeserverTestCase):
 
         # Blow away caches (supported room versions can only change due to a restart).
         self.store.get_rooms_for_user_with_stream_ordering.invalidate_all()
+        self.store.get_rooms_for_user.invalidate_all()
         self.get_success(self.store._get_event_cache.clear())
         self.store._event_ref.clear()
 
diff --git a/tests/push/test_push_rule_evaluator.py b/tests/push/test_push_rule_evaluator.py
index 718f489577..8804f0e0d3 100644
--- a/tests/push/test_push_rule_evaluator.py
+++ b/tests/push/test_push_rule_evaluator.py
@@ -19,16 +19,18 @@ import frozendict
 from twisted.test.proto_helpers import MemoryReactor
 
 import synapse.rest.admin
-from synapse.api.constants import EventTypes, Membership
+from synapse.api.constants import EventTypes, HistoryVisibility, Membership
 from synapse.api.room_versions import RoomVersions
 from synapse.appservice import ApplicationService
 from synapse.events import FrozenEvent
-from synapse.push import push_rule_evaluator
-from synapse.push.push_rule_evaluator import PushRuleEvaluatorForEvent
+from synapse.push.bulk_push_rule_evaluator import _flatten_dict
+from synapse.push.httppusher import tweaks_for_actions
+from synapse.rest import admin
 from synapse.rest.client import login, register, room
 from synapse.server import HomeServer
 from synapse.storage.databases.main.appservice import _make_exclusive_regex
-from synapse.types import JsonDict
+from synapse.synapse_rust.push import PushRuleEvaluator
+from synapse.types import JsonDict, UserID
 from synapse.util import Clock
 
 from tests import unittest
@@ -41,7 +43,7 @@ class PushRuleEvaluatorTestCase(unittest.TestCase):
         content: JsonDict,
         relations: Optional[Dict[str, Set[Tuple[str, str]]]] = None,
         relations_match_enabled: bool = False,
-    ) -> PushRuleEvaluatorForEvent:
+    ) -> PushRuleEvaluator:
         event = FrozenEvent(
             {
                 "event_id": "$event_id",
@@ -56,12 +58,12 @@ class PushRuleEvaluatorTestCase(unittest.TestCase):
         room_member_count = 0
         sender_power_level = 0
         power_levels: Dict[str, Union[int, Dict[str, int]]] = {}
-        return PushRuleEvaluatorForEvent(
-            event,
+        return PushRuleEvaluator(
+            _flatten_dict(event),
             room_member_count,
             sender_power_level,
-            power_levels,
-            relations or set(),
+            power_levels.get("notifications", {}),
+            relations or {},
             relations_match_enabled,
         )
 
@@ -293,7 +295,7 @@ class PushRuleEvaluatorTestCase(unittest.TestCase):
         ]
 
         self.assertEqual(
-            push_rule_evaluator.tweaks_for_actions(actions),
+            tweaks_for_actions(actions),
             {"sound": "default", "highlight": True},
         )
 
@@ -304,9 +306,6 @@ class PushRuleEvaluatorTestCase(unittest.TestCase):
         evaluator = self._get_evaluator(
             {}, {"m.annotation": {("@user:test", "m.reaction")}}
         )
-        condition = {"kind": "relation_match"}
-        # Oddly, an unknown condition always matches.
-        self.assertTrue(evaluator.matches(condition, "@user:test", "foo"))
 
         # A push rule evaluator with the experimental rule enabled.
         evaluator = self._get_evaluator(
@@ -439,3 +438,80 @@ class TestBulkPushRuleEvaluator(unittest.HomeserverTestCase):
         )
 
         self.assertEqual(len(users_with_push_actions), 0)
+
+
+class BulkPushRuleEvaluatorTestCase(unittest.HomeserverTestCase):
+    servlets = [
+        admin.register_servlets,
+        login.register_servlets,
+        room.register_servlets,
+    ]
+
+    def prepare(
+        self, reactor: MemoryReactor, clock: Clock, homeserver: HomeServer
+    ) -> None:
+        self.main_store = homeserver.get_datastores().main
+
+        self.user_id1 = self.register_user("user1", "password")
+        self.tok1 = self.login(self.user_id1, "password")
+        self.user_id2 = self.register_user("user2", "password")
+        self.tok2 = self.login(self.user_id2, "password")
+
+        self.room_id = self.helper.create_room_as(tok=self.tok1)
+
+        # We want to test history visibility works correctly.
+        self.helper.send_state(
+            self.room_id,
+            EventTypes.RoomHistoryVisibility,
+            {"history_visibility": HistoryVisibility.JOINED},
+            tok=self.tok1,
+        )
+
+    def get_notif_count(self, user_id: str) -> int:
+        return self.get_success(
+            self.main_store.db_pool.simple_select_one_onecol(
+                table="event_push_actions",
+                keyvalues={"user_id": user_id},
+                retcol="COALESCE(SUM(notif), 0)",
+                desc="get_staging_notif_count",
+            )
+        )
+
+    def test_plain_message(self) -> None:
+        """Test that sending a normal message in a room will trigger a
+        notification
+        """
+
+        # Have user2 join the room and cle
+        self.helper.join(self.room_id, self.user_id2, tok=self.tok2)
+
+        # They start off with no notifications, but get them when messages are
+        # sent.
+        self.assertEqual(self.get_notif_count(self.user_id2), 0)
+
+        user1 = UserID.from_string(self.user_id1)
+        self.create_and_send_event(self.room_id, user1)
+
+        self.assertEqual(self.get_notif_count(self.user_id2), 1)
+
+    def test_delayed_message(self) -> None:
+        """Test that a delayed message that was from before a user joined
+        doesn't cause a notification for the joined user.
+        """
+        user1 = UserID.from_string(self.user_id1)
+
+        # Send a message before user2 joins
+        event_id1 = self.create_and_send_event(self.room_id, user1)
+
+        # Have user2 join the room
+        self.helper.join(self.room_id, self.user_id2, tok=self.tok2)
+
+        # They start off with no notifications
+        self.assertEqual(self.get_notif_count(self.user_id2), 0)
+
+        # Send another message that references the event before the join to
+        # simulate a "delayed" event
+        self.create_and_send_event(self.room_id, user1, prev_event_ids=[event_id1])
+
+        # user2 should not be notified about it, because they can't see it.
+        self.assertEqual(self.get_notif_count(self.user_id2), 0)
diff --git a/tests/rest/admin/test_user.py b/tests/rest/admin/test_user.py
index 1847e6ad6b..4c1ce33463 100644
--- a/tests/rest/admin/test_user.py
+++ b/tests/rest/admin/test_user.py
@@ -25,10 +25,10 @@ from parameterized import parameterized, parameterized_class
 from twisted.test.proto_helpers import MemoryReactor
 
 import synapse.rest.admin
-from synapse.api.constants import UserTypes
+from synapse.api.constants import ApprovalNoticeMedium, LoginType, UserTypes
 from synapse.api.errors import Codes, HttpResponseException, ResourceLimitError
 from synapse.api.room_versions import RoomVersions
-from synapse.rest.client import devices, login, logout, profile, room, sync
+from synapse.rest.client import devices, login, logout, profile, register, room, sync
 from synapse.rest.media.v1.filepath import MediaFilePaths
 from synapse.server import HomeServer
 from synapse.types import JsonDict, UserID
@@ -578,6 +578,16 @@ class UsersListTestCase(unittest.HomeserverTestCase):
         _search_test(None, "foo", "user_id")
         _search_test(None, "bar", "user_id")
 
+    @override_config(
+        {
+            "experimental_features": {
+                "msc3866": {
+                    "enabled": True,
+                    "require_approval_for_new_accounts": True,
+                }
+            }
+        }
+    )
     def test_invalid_parameter(self) -> None:
         """
         If parameters are invalid, an error is returned.
@@ -623,6 +633,16 @@ class UsersListTestCase(unittest.HomeserverTestCase):
         self.assertEqual(400, channel.code, msg=channel.json_body)
         self.assertEqual(Codes.INVALID_PARAM, channel.json_body["errcode"])
 
+        # invalid approved
+        channel = self.make_request(
+            "GET",
+            self.url + "?approved=not_bool",
+            access_token=self.admin_user_tok,
+        )
+
+        self.assertEqual(400, channel.code, msg=channel.json_body)
+        self.assertEqual(Codes.INVALID_PARAM, channel.json_body["errcode"])
+
         # unkown order_by
         channel = self.make_request(
             "GET",
@@ -841,6 +861,69 @@ class UsersListTestCase(unittest.HomeserverTestCase):
         self._order_test([self.admin_user, user1, user2], "creation_ts", "f")
         self._order_test([user2, user1, self.admin_user], "creation_ts", "b")
 
+    @override_config(
+        {
+            "experimental_features": {
+                "msc3866": {
+                    "enabled": True,
+                    "require_approval_for_new_accounts": True,
+                }
+            }
+        }
+    )
+    def test_filter_out_approved(self) -> None:
+        """Tests that the endpoint can filter out approved users."""
+        # Create our users.
+        self._create_users(2)
+
+        # Get the list of users.
+        channel = self.make_request(
+            "GET",
+            self.url,
+            access_token=self.admin_user_tok,
+        )
+        self.assertEqual(200, channel.code, channel.result)
+
+        # Exclude the admin, because we don't want to accidentally un-approve the admin.
+        non_admin_user_ids = [
+            user["name"]
+            for user in channel.json_body["users"]
+            if user["name"] != self.admin_user
+        ]
+
+        self.assertEqual(2, len(non_admin_user_ids), non_admin_user_ids)
+
+        # Select a user and un-approve them. We do this rather than the other way around
+        # because, since these users are created by an admin, we consider them already
+        # approved.
+        not_approved_user = non_admin_user_ids[0]
+
+        channel = self.make_request(
+            "PUT",
+            f"/_synapse/admin/v2/users/{not_approved_user}",
+            {"approved": False},
+            access_token=self.admin_user_tok,
+        )
+        self.assertEqual(200, channel.code, channel.result)
+
+        # Now get the list of users again, this time filtering out approved users.
+        channel = self.make_request(
+            "GET",
+            self.url + "?approved=false",
+            access_token=self.admin_user_tok,
+        )
+        self.assertEqual(200, channel.code, channel.result)
+
+        non_admin_user_ids = [
+            user["name"]
+            for user in channel.json_body["users"]
+            if user["name"] != self.admin_user
+        ]
+
+        # We should only have our unapproved user now.
+        self.assertEqual(1, len(non_admin_user_ids), non_admin_user_ids)
+        self.assertEqual(not_approved_user, non_admin_user_ids[0])
+
     def _order_test(
         self,
         expected_user_list: List[str],
@@ -1272,6 +1355,7 @@ class UserRestTestCase(unittest.HomeserverTestCase):
         synapse.rest.admin.register_servlets,
         login.register_servlets,
         sync.register_servlets,
+        register.register_servlets,
     ]
 
     def prepare(self, reactor: MemoryReactor, clock: Clock, hs: HomeServer) -> None:
@@ -2536,6 +2620,104 @@ class UserRestTestCase(unittest.HomeserverTestCase):
         # Ensure they're still alive
         self.assertEqual(0, channel.json_body["deactivated"])
 
+    @override_config(
+        {
+            "experimental_features": {
+                "msc3866": {
+                    "enabled": True,
+                    "require_approval_for_new_accounts": True,
+                }
+            }
+        }
+    )
+    def test_approve_account(self) -> None:
+        """Tests that approving an account correctly sets the approved flag for the user."""
+        url = self.url_prefix % "@bob:test"
+
+        # Create the user using the client-server API since otherwise the user will be
+        # marked as approved automatically.
+        channel = self.make_request(
+            "POST",
+            "register",
+            {
+                "username": "bob",
+                "password": "test",
+                "auth": {"type": LoginType.DUMMY},
+            },
+        )
+        self.assertEqual(403, channel.code, channel.result)
+        self.assertEqual(Codes.USER_AWAITING_APPROVAL, channel.json_body["errcode"])
+        self.assertEqual(
+            ApprovalNoticeMedium.NONE, channel.json_body["approval_notice_medium"]
+        )
+
+        # Get user
+        channel = self.make_request(
+            "GET",
+            url,
+            access_token=self.admin_user_tok,
+        )
+
+        self.assertEqual(200, channel.code, msg=channel.json_body)
+        self.assertIs(False, channel.json_body["approved"])
+
+        # Approve user
+        channel = self.make_request(
+            "PUT",
+            url,
+            access_token=self.admin_user_tok,
+            content={"approved": True},
+        )
+
+        self.assertEqual(200, channel.code, msg=channel.json_body)
+        self.assertIs(True, channel.json_body["approved"])
+
+        # Check that the user is now approved
+        channel = self.make_request(
+            "GET",
+            url,
+            access_token=self.admin_user_tok,
+        )
+
+        self.assertEqual(200, channel.code, msg=channel.json_body)
+        self.assertIs(True, channel.json_body["approved"])
+
+    @override_config(
+        {
+            "experimental_features": {
+                "msc3866": {
+                    "enabled": True,
+                    "require_approval_for_new_accounts": True,
+                }
+            }
+        }
+    )
+    def test_register_approved(self) -> None:
+        url = self.url_prefix % "@bob:test"
+
+        # Create user
+        channel = self.make_request(
+            "PUT",
+            url,
+            access_token=self.admin_user_tok,
+            content={"password": "abc123", "approved": True},
+        )
+
+        self.assertEqual(201, channel.code, msg=channel.json_body)
+        self.assertEqual("@bob:test", channel.json_body["name"])
+        self.assertEqual(1, channel.json_body["approved"])
+
+        # Get user
+        channel = self.make_request(
+            "GET",
+            url,
+            access_token=self.admin_user_tok,
+        )
+
+        self.assertEqual(200, channel.code, msg=channel.json_body)
+        self.assertEqual("@bob:test", channel.json_body["name"])
+        self.assertEqual(1, channel.json_body["approved"])
+
     def _is_erased(self, user_id: str, expect: bool) -> None:
         """Assert that the user is erased or not"""
         d = self.store.is_user_erased(user_id)
diff --git a/tests/rest/client/test_auth.py b/tests/rest/client/test_auth.py
index 05355c7fb6..090cef5216 100644
--- a/tests/rest/client/test_auth.py
+++ b/tests/rest/client/test_auth.py
@@ -20,7 +20,8 @@ from twisted.test.proto_helpers import MemoryReactor
 from twisted.web.resource import Resource
 
 import synapse.rest.admin
-from synapse.api.constants import LoginType
+from synapse.api.constants import ApprovalNoticeMedium, LoginType
+from synapse.api.errors import Codes
 from synapse.handlers.ui_auth.checkers import UserInteractiveAuthChecker
 from synapse.rest.client import account, auth, devices, login, logout, register
 from synapse.rest.synapse.client import build_synapse_client_resource_tree
@@ -567,6 +568,36 @@ class UIAuthTests(unittest.HomeserverTestCase):
             body={"auth": {"session": session_id}},
         )
 
+    @skip_unless(HAS_OIDC, "requires OIDC")
+    @override_config(
+        {
+            "oidc_config": TEST_OIDC_CONFIG,
+            "experimental_features": {
+                "msc3866": {
+                    "enabled": True,
+                    "require_approval_for_new_accounts": True,
+                }
+            },
+        }
+    )
+    def test_sso_not_approved(self) -> None:
+        """Tests that if we register a user via SSO while requiring approval for new
+        accounts, we still raise the correct error before logging the user in.
+        """
+        login_resp = self.helper.login_via_oidc("username", expected_status=403)
+
+        self.assertEqual(login_resp["errcode"], Codes.USER_AWAITING_APPROVAL)
+        self.assertEqual(
+            ApprovalNoticeMedium.NONE, login_resp["approval_notice_medium"]
+        )
+
+        # Check that we didn't register a device for the user during the login attempt.
+        devices = self.get_success(
+            self.hs.get_datastores().main.get_devices_by_user("@username:test")
+        )
+
+        self.assertEqual(len(devices), 0)
+
 
 class RefreshAuthTests(unittest.HomeserverTestCase):
     servlets = [
diff --git a/tests/rest/client/test_login.py b/tests/rest/client/test_login.py
index e2a4d98275..e801ba8c8b 100644
--- a/tests/rest/client/test_login.py
+++ b/tests/rest/client/test_login.py
@@ -23,6 +23,8 @@ from twisted.test.proto_helpers import MemoryReactor
 from twisted.web.resource import Resource
 
 import synapse.rest.admin
+from synapse.api.constants import ApprovalNoticeMedium, LoginType
+from synapse.api.errors import Codes
 from synapse.appservice import ApplicationService
 from synapse.rest.client import devices, login, logout, register
 from synapse.rest.client.account import WhoamiRestServlet
@@ -94,6 +96,7 @@ class LoginRestServletTestCase(unittest.HomeserverTestCase):
         logout.register_servlets,
         devices.register_servlets,
         lambda hs, http_server: WhoamiRestServlet(hs).register(http_server),
+        register.register_servlets,
     ]
 
     def make_homeserver(self, reactor: MemoryReactor, clock: Clock) -> HomeServer:
@@ -406,6 +409,44 @@ class LoginRestServletTestCase(unittest.HomeserverTestCase):
         self.assertEqual(channel.code, 400)
         self.assertEqual(channel.json_body["errcode"], "M_INVALID_PARAM")
 
+    @override_config(
+        {
+            "experimental_features": {
+                "msc3866": {
+                    "enabled": True,
+                    "require_approval_for_new_accounts": True,
+                }
+            }
+        }
+    )
+    def test_require_approval(self) -> None:
+        channel = self.make_request(
+            "POST",
+            "register",
+            {
+                "username": "kermit",
+                "password": "monkey",
+                "auth": {"type": LoginType.DUMMY},
+            },
+        )
+        self.assertEqual(403, channel.code, channel.result)
+        self.assertEqual(Codes.USER_AWAITING_APPROVAL, channel.json_body["errcode"])
+        self.assertEqual(
+            ApprovalNoticeMedium.NONE, channel.json_body["approval_notice_medium"]
+        )
+
+        params = {
+            "type": LoginType.PASSWORD,
+            "identifier": {"type": "m.id.user", "user": "kermit"},
+            "password": "monkey",
+        }
+        channel = self.make_request("POST", LOGIN_URL, params)
+        self.assertEqual(403, channel.code, channel.result)
+        self.assertEqual(Codes.USER_AWAITING_APPROVAL, channel.json_body["errcode"])
+        self.assertEqual(
+            ApprovalNoticeMedium.NONE, channel.json_body["approval_notice_medium"]
+        )
+
 
 @skip_unless(has_saml2 and HAS_OIDC, "Requires SAML2 and OIDC")
 class MultiSSOTestCase(unittest.HomeserverTestCase):
diff --git a/tests/rest/client/test_login_token_request.py b/tests/rest/client/test_login_token_request.py
index d5bb16c98d..c2e1e08811 100644
--- a/tests/rest/client/test_login_token_request.py
+++ b/tests/rest/client/test_login_token_request.py
@@ -22,6 +22,8 @@ from synapse.util import Clock
 from tests import unittest
 from tests.unittest import override_config
 
+endpoint = "/_matrix/client/unstable/org.matrix.msc3882/login/token"
+
 
 class LoginTokenRequestServletTestCase(unittest.HomeserverTestCase):
 
@@ -45,18 +47,18 @@ class LoginTokenRequestServletTestCase(unittest.HomeserverTestCase):
         self.password = "password"
 
     def test_disabled(self) -> None:
-        channel = self.make_request("POST", "/login/token", {}, access_token=None)
+        channel = self.make_request("POST", endpoint, {}, access_token=None)
         self.assertEqual(channel.code, 400)
 
         self.register_user(self.user, self.password)
         token = self.login(self.user, self.password)
 
-        channel = self.make_request("POST", "/login/token", {}, access_token=token)
+        channel = self.make_request("POST", endpoint, {}, access_token=token)
         self.assertEqual(channel.code, 400)
 
     @override_config({"experimental_features": {"msc3882_enabled": True}})
     def test_require_auth(self) -> None:
-        channel = self.make_request("POST", "/login/token", {}, access_token=None)
+        channel = self.make_request("POST", endpoint, {}, access_token=None)
         self.assertEqual(channel.code, 401)
 
     @override_config({"experimental_features": {"msc3882_enabled": True}})
@@ -64,7 +66,7 @@ class LoginTokenRequestServletTestCase(unittest.HomeserverTestCase):
         user_id = self.register_user(self.user, self.password)
         token = self.login(self.user, self.password)
 
-        channel = self.make_request("POST", "/login/token", {}, access_token=token)
+        channel = self.make_request("POST", endpoint, {}, access_token=token)
         self.assertEqual(channel.code, 401)
         self.assertIn({"stages": ["m.login.password"]}, channel.json_body["flows"])
 
@@ -79,7 +81,7 @@ class LoginTokenRequestServletTestCase(unittest.HomeserverTestCase):
             },
         }
 
-        channel = self.make_request("POST", "/login/token", uia, access_token=token)
+        channel = self.make_request("POST", endpoint, uia, access_token=token)
         self.assertEqual(channel.code, 200)
         self.assertEqual(channel.json_body["expires_in"], 300)
 
@@ -100,7 +102,7 @@ class LoginTokenRequestServletTestCase(unittest.HomeserverTestCase):
         user_id = self.register_user(self.user, self.password)
         token = self.login(self.user, self.password)
 
-        channel = self.make_request("POST", "/login/token", {}, access_token=token)
+        channel = self.make_request("POST", endpoint, {}, access_token=token)
         self.assertEqual(channel.code, 200)
         self.assertEqual(channel.json_body["expires_in"], 300)
 
@@ -127,6 +129,6 @@ class LoginTokenRequestServletTestCase(unittest.HomeserverTestCase):
         self.register_user(self.user, self.password)
         token = self.login(self.user, self.password)
 
-        channel = self.make_request("POST", "/login/token", {}, access_token=token)
+        channel = self.make_request("POST", endpoint, {}, access_token=token)
         self.assertEqual(channel.code, 200)
         self.assertEqual(channel.json_body["expires_in"], 15)
diff --git a/tests/rest/client/test_register.py b/tests/rest/client/test_register.py
index b781875d52..11cf3939d8 100644
--- a/tests/rest/client/test_register.py
+++ b/tests/rest/client/test_register.py
@@ -22,7 +22,11 @@ import pkg_resources
 from twisted.test.proto_helpers import MemoryReactor
 
 import synapse.rest.admin
-from synapse.api.constants import APP_SERVICE_REGISTRATION_TYPE, LoginType
+from synapse.api.constants import (
+    APP_SERVICE_REGISTRATION_TYPE,
+    ApprovalNoticeMedium,
+    LoginType,
+)
 from synapse.api.errors import Codes
 from synapse.appservice import ApplicationService
 from synapse.rest.client import account, account_validity, login, logout, register, sync
@@ -765,6 +769,32 @@ class RegisterRestServletTestCase(unittest.HomeserverTestCase):
         self.assertEqual(channel.code, 400, channel.json_body)
         self.assertEqual(channel.json_body["errcode"], Codes.USER_IN_USE)
 
+    @override_config(
+        {
+            "experimental_features": {
+                "msc3866": {
+                    "enabled": True,
+                    "require_approval_for_new_accounts": True,
+                }
+            }
+        }
+    )
+    def test_require_approval(self) -> None:
+        channel = self.make_request(
+            "POST",
+            "register",
+            {
+                "username": "kermit",
+                "password": "monkey",
+                "auth": {"type": LoginType.DUMMY},
+            },
+        )
+        self.assertEqual(403, channel.code, channel.result)
+        self.assertEqual(Codes.USER_AWAITING_APPROVAL, channel.json_body["errcode"])
+        self.assertEqual(
+            ApprovalNoticeMedium.NONE, channel.json_body["approval_notice_medium"]
+        )
+
 
 class AccountValidityTestCase(unittest.HomeserverTestCase):
 
diff --git a/tests/rest/client/test_relations.py b/tests/rest/client/test_relations.py
index d33e34d829..fef3b72d76 100644
--- a/tests/rest/client/test_relations.py
+++ b/tests/rest/client/test_relations.py
@@ -728,7 +728,6 @@ class RelationsTestCase(BaseRelationsTestCase):
 
 
 class RelationPaginationTestCase(BaseRelationsTestCase):
-    @unittest.override_config({"experimental_features": {"msc3715_enabled": True}})
     def test_basic_paginate_relations(self) -> None:
         """Tests that calling pagination API correctly the latest relations."""
         channel = self._send_relation(RelationTypes.ANNOTATION, "m.reaction", "a")
@@ -771,7 +770,7 @@ class RelationPaginationTestCase(BaseRelationsTestCase):
         channel = self.make_request(
             "GET",
             f"/_matrix/client/v1/rooms/{self.room}/relations"
-            f"/{self.parent_id}?limit=1&org.matrix.msc3715.dir=f",
+            f"/{self.parent_id}?limit=1&dir=f",
             access_token=self.user_token,
         )
         self.assertEqual(200, channel.code, channel.json_body)
@@ -788,7 +787,6 @@ class RelationPaginationTestCase(BaseRelationsTestCase):
             channel.json_body["chunk"][0],
         )
 
-    @unittest.override_config({"experimental_features": {"msc3715_enabled": True}})
     def test_repeated_paginate_relations(self) -> None:
         """Test that if we paginate using a limit and tokens then we get the
         expected events.
@@ -838,7 +836,7 @@ class RelationPaginationTestCase(BaseRelationsTestCase):
 
             channel = self.make_request(
                 "GET",
-                f"/_matrix/client/v1/rooms/{self.room}/relations/{self.parent_id}?org.matrix.msc3715.dir=f&limit=3{from_token}",
+                f"/_matrix/client/v1/rooms/{self.room}/relations/{self.parent_id}?dir=f&limit=3{from_token}",
                 access_token=self.user_token,
             )
             self.assertEqual(200, channel.code, channel.json_body)
diff --git a/tests/rest/client/test_rooms.py b/tests/rest/client/test_rooms.py
index c7eb88d33f..7f8cf4fab0 100644
--- a/tests/rest/client/test_rooms.py
+++ b/tests/rest/client/test_rooms.py
@@ -710,7 +710,7 @@ class RoomsCreateTestCase(RoomBase):
         self.assertEqual(HTTPStatus.OK, channel.code, channel.result)
         self.assertTrue("room_id" in channel.json_body)
         assert channel.resource_usage is not None
-        self.assertEqual(44, channel.resource_usage.db_txn_count)
+        self.assertEqual(34, channel.resource_usage.db_txn_count)
 
     def test_post_room_initial_state(self) -> None:
         # POST with initial_state config key, expect new room id
@@ -723,7 +723,7 @@ class RoomsCreateTestCase(RoomBase):
         self.assertEqual(HTTPStatus.OK, channel.code, channel.result)
         self.assertTrue("room_id" in channel.json_body)
         assert channel.resource_usage is not None
-        self.assertEqual(50, channel.resource_usage.db_txn_count)
+        self.assertEqual(37, channel.resource_usage.db_txn_count)
 
     def test_post_room_visibility_key(self) -> None:
         # POST with visibility config key, expect new room id
diff --git a/tests/rest/client/utils.py b/tests/rest/client/utils.py
index dd26145bf8..c249a42bb6 100644
--- a/tests/rest/client/utils.py
+++ b/tests/rest/client/utils.py
@@ -543,8 +543,12 @@ class RestHelper:
 
         return channel.json_body
 
-    def login_via_oidc(self, remote_user_id: str) -> JsonDict:
-        """Log in (as a new user) via OIDC
+    def login_via_oidc(
+        self,
+        remote_user_id: str,
+        expected_status: int = 200,
+    ) -> JsonDict:
+        """Log in via OIDC
 
         Returns the result of the final token login.
 
@@ -578,7 +582,9 @@ class RestHelper:
             "/login",
             content={"type": "m.login.token", "token": login_token},
         )
-        assert channel.code == HTTPStatus.OK
+        assert (
+            channel.code == expected_status
+        ), f"unexpected status in response: {channel.code}"
         return channel.json_body
 
     def auth_via_oidc(
diff --git a/tests/storage/databases/main/test_events_worker.py b/tests/storage/databases/main/test_events_worker.py
index 67401272ac..32a798d74b 100644
--- a/tests/storage/databases/main/test_events_worker.py
+++ b/tests/storage/databases/main/test_events_worker.py
@@ -35,66 +35,45 @@ from synapse.util import Clock
 from synapse.util.async_helpers import yieldable_gather_results
 
 from tests import unittest
+from tests.test_utils.event_injection import create_event, inject_event
 
 
 class HaveSeenEventsTestCase(unittest.HomeserverTestCase):
+    servlets = [
+        admin.register_servlets,
+        room.register_servlets,
+        login.register_servlets,
+    ]
+
     def prepare(self, reactor, clock, hs):
+        self.hs = hs
         self.store: EventsWorkerStore = hs.get_datastores().main
 
-        # insert some test data
-        for rid in ("room1", "room2"):
-            self.get_success(
-                self.store.db_pool.simple_insert(
-                    "rooms",
-                    {"room_id": rid, "room_version": 4},
-                )
-            )
+        self.user = self.register_user("user", "pass")
+        self.token = self.login(self.user, "pass")
+        self.room_id = self.helper.create_room_as(self.user, tok=self.token)
 
         self.event_ids: List[str] = []
-        for idx, rid in enumerate(
-            (
-                "room1",
-                "room1",
-                "room1",
-                "room2",
-            )
-        ):
-            event_json = {"type": f"test {idx}", "room_id": rid}
-            event = make_event_from_dict(event_json, room_version=RoomVersions.V4)
-            event_id = event.event_id
-
-            self.get_success(
-                self.store.db_pool.simple_insert(
-                    "events",
-                    {
-                        "event_id": event_id,
-                        "room_id": rid,
-                        "topological_ordering": idx,
-                        "stream_ordering": idx,
-                        "type": event.type,
-                        "processed": True,
-                        "outlier": False,
-                    },
+        for i in range(3):
+            event = self.get_success(
+                inject_event(
+                    hs,
+                    room_version=RoomVersions.V7.identifier,
+                    room_id=self.room_id,
+                    sender=self.user,
+                    type="test_event_type",
+                    content={"body": f"foobarbaz{i}"},
                 )
             )
-            self.get_success(
-                self.store.db_pool.simple_insert(
-                    "event_json",
-                    {
-                        "event_id": event_id,
-                        "room_id": rid,
-                        "json": json.dumps(event_json),
-                        "internal_metadata": "{}",
-                        "format_version": 3,
-                    },
-                )
-            )
-            self.event_ids.append(event_id)
+
+            self.event_ids.append(event.event_id)
 
     def test_simple(self):
         with LoggingContext(name="test") as ctx:
             res = self.get_success(
-                self.store.have_seen_events("room1", [self.event_ids[0], "event19"])
+                self.store.have_seen_events(
+                    self.room_id, [self.event_ids[0], "eventdoesnotexist"]
+                )
             )
             self.assertEqual(res, {self.event_ids[0]})
 
@@ -104,7 +83,9 @@ class HaveSeenEventsTestCase(unittest.HomeserverTestCase):
         # a second lookup of the same events should cause no queries
         with LoggingContext(name="test") as ctx:
             res = self.get_success(
-                self.store.have_seen_events("room1", [self.event_ids[0], "event19"])
+                self.store.have_seen_events(
+                    self.room_id, [self.event_ids[0], "eventdoesnotexist"]
+                )
             )
             self.assertEqual(res, {self.event_ids[0]})
             self.assertEqual(ctx.get_resource_usage().db_txn_count, 0)
@@ -116,11 +97,86 @@ class HaveSeenEventsTestCase(unittest.HomeserverTestCase):
         # looking it up should now cause no db hits
         with LoggingContext(name="test") as ctx:
             res = self.get_success(
-                self.store.have_seen_events("room1", [self.event_ids[0]])
+                self.store.have_seen_events(self.room_id, [self.event_ids[0]])
             )
             self.assertEqual(res, {self.event_ids[0]})
             self.assertEqual(ctx.get_resource_usage().db_txn_count, 0)
 
+    def test_persisting_event_invalidates_cache(self):
+        """
+        Test to make sure that the `have_seen_event` cache
+        is invalidated after we persist an event and returns
+        the updated value.
+        """
+        event, event_context = self.get_success(
+            create_event(
+                self.hs,
+                room_id=self.room_id,
+                sender=self.user,
+                type="test_event_type",
+                content={"body": "garply"},
+            )
+        )
+
+        with LoggingContext(name="test") as ctx:
+            # First, check `have_seen_event` for an event we have not seen yet
+            # to prime the cache with a `false` value.
+            res = self.get_success(
+                self.store.have_seen_events(event.room_id, [event.event_id])
+            )
+            self.assertEqual(res, set())
+
+            # That should result in a single db query to lookup
+            self.assertEqual(ctx.get_resource_usage().db_txn_count, 1)
+
+        # Persist the event which should invalidate or prefill the
+        # `have_seen_event` cache so we don't return stale values.
+        persistence = self.hs.get_storage_controllers().persistence
+        self.get_success(
+            persistence.persist_event(
+                event,
+                event_context,
+            )
+        )
+
+        with LoggingContext(name="test") as ctx:
+            # Check `have_seen_event` again and we should see the updated fact
+            # that we have now seen the event after persisting it.
+            res = self.get_success(
+                self.store.have_seen_events(event.room_id, [event.event_id])
+            )
+            self.assertEqual(res, {event.event_id})
+
+            # That should result in a single db query to lookup
+            self.assertEqual(ctx.get_resource_usage().db_txn_count, 1)
+
+    def test_invalidate_cache_by_room_id(self):
+        """
+        Test to make sure that all events associated with the given `(room_id,)`
+        are invalidated in the `have_seen_event` cache.
+        """
+        with LoggingContext(name="test") as ctx:
+            # Prime the cache with some values
+            res = self.get_success(
+                self.store.have_seen_events(self.room_id, self.event_ids)
+            )
+            self.assertEqual(res, set(self.event_ids))
+
+            # That should result in a single db query to lookup
+            self.assertEqual(ctx.get_resource_usage().db_txn_count, 1)
+
+        # Clear the cache with any events associated with the `room_id`
+        self.store.have_seen_event.invalidate((self.room_id,))
+
+        with LoggingContext(name="test") as ctx:
+            res = self.get_success(
+                self.store.have_seen_events(self.room_id, self.event_ids)
+            )
+            self.assertEqual(res, set(self.event_ids))
+
+            # Since we cleared the cache, it should result in another db query to lookup
+            self.assertEqual(ctx.get_resource_usage().db_txn_count, 1)
+
 
 class EventCacheTestCase(unittest.HomeserverTestCase):
     """Test that the various layers of event cache works."""
diff --git a/tests/storage/test_event_chain.py b/tests/storage/test_event_chain.py
index a0ce077a99..de9f4af2de 100644
--- a/tests/storage/test_event_chain.py
+++ b/tests/storage/test_event_chain.py
@@ -531,7 +531,9 @@ class EventChainBackgroundUpdateTestCase(HomeserverTestCase):
             )
         )
         self.get_success(
-            event_handler.handle_new_client_event(self.requester, event, context)
+            event_handler.handle_new_client_event(
+                self.requester, events_and_context=[(event, context)]
+            )
         )
         state1 = set(self.get_success(context.get_current_state_ids()).values())
 
@@ -549,7 +551,9 @@ class EventChainBackgroundUpdateTestCase(HomeserverTestCase):
             )
         )
         self.get_success(
-            event_handler.handle_new_client_event(self.requester, event, context)
+            event_handler.handle_new_client_event(
+                self.requester, events_and_context=[(event, context)]
+            )
         )
         state2 = set(self.get_success(context.get_current_state_ids()).values())
 
diff --git a/tests/storage/test_event_federation.py b/tests/storage/test_event_federation.py
index 85739c464e..59b8910907 100644
--- a/tests/storage/test_event_federation.py
+++ b/tests/storage/test_event_federation.py
@@ -754,18 +754,28 @@ class EventFederationWorkerStoreTestCase(tests.unittest.HomeserverTestCase):
 
     def test_get_backfill_points_in_room(self):
         """
-        Test to make sure we get some backfill points
+        Test to make sure only backfill points that are older and come before
+        the `current_depth` are returned.
         """
         setup_info = self._setup_room_for_backfill_tests()
         room_id = setup_info.room_id
+        depth_map = setup_info.depth_map
 
+        # Try at "B"
         backfill_points = self.get_success(
-            self.store.get_backfill_points_in_room(room_id)
+            self.store.get_backfill_points_in_room(room_id, depth_map["B"], limit=100)
         )
         backfill_event_ids = [backfill_point[0] for backfill_point in backfill_points]
-        self.assertListEqual(
-            backfill_event_ids, ["b6", "b5", "b4", "2", "b3", "b2", "b1"]
+        self.assertEqual(backfill_event_ids, ["b6", "b5", "b4", "2", "b3", "b2", "b1"])
+
+        # Try at "A"
+        backfill_points = self.get_success(
+            self.store.get_backfill_points_in_room(room_id, depth_map["A"], limit=100)
         )
+        backfill_event_ids = [backfill_point[0] for backfill_point in backfill_points]
+        # Event "2" has a depth of 2 but is not included here because we only
+        # know the approximate depth of 5 from our event "3".
+        self.assertListEqual(backfill_event_ids, ["b3", "b2", "b1"])
 
     def test_get_backfill_points_in_room_excludes_events_we_have_attempted(
         self,
@@ -776,6 +786,7 @@ class EventFederationWorkerStoreTestCase(tests.unittest.HomeserverTestCase):
         """
         setup_info = self._setup_room_for_backfill_tests()
         room_id = setup_info.room_id
+        depth_map = setup_info.depth_map
 
         # Record some attempts to backfill these events which will make
         # `get_backfill_points_in_room` exclude them because we
@@ -795,12 +806,13 @@ class EventFederationWorkerStoreTestCase(tests.unittest.HomeserverTestCase):
 
         # No time has passed since we attempted to backfill ^
 
+        # Try at "B"
         backfill_points = self.get_success(
-            self.store.get_backfill_points_in_room(room_id)
+            self.store.get_backfill_points_in_room(room_id, depth_map["B"], limit=100)
         )
         backfill_event_ids = [backfill_point[0] for backfill_point in backfill_points]
         # Only the backfill points that we didn't record earlier exist here.
-        self.assertListEqual(backfill_event_ids, ["b6", "2", "b1"])
+        self.assertEqual(backfill_event_ids, ["b6", "2", "b1"])
 
     def test_get_backfill_points_in_room_attempted_event_retry_after_backoff_duration(
         self,
@@ -812,6 +824,7 @@ class EventFederationWorkerStoreTestCase(tests.unittest.HomeserverTestCase):
         """
         setup_info = self._setup_room_for_backfill_tests()
         room_id = setup_info.room_id
+        depth_map = setup_info.depth_map
 
         # Record some attempts to backfill these events which will make
         # `get_backfill_points_in_room` exclude them because we
@@ -839,27 +852,66 @@ class EventFederationWorkerStoreTestCase(tests.unittest.HomeserverTestCase):
         # visible regardless.
         self.reactor.advance(datetime.timedelta(hours=2).total_seconds())
 
-        # Make sure that "b1" is not in the list because we've
+        # Try at "A" and make sure that "b1" is not in the list because we've
         # already attempted many times
         backfill_points = self.get_success(
-            self.store.get_backfill_points_in_room(room_id)
+            self.store.get_backfill_points_in_room(room_id, depth_map["A"], limit=100)
         )
         backfill_event_ids = [backfill_point[0] for backfill_point in backfill_points]
-        self.assertListEqual(backfill_event_ids, ["b6", "b5", "b4", "2", "b3", "b2"])
+        self.assertEqual(backfill_event_ids, ["b3", "b2"])
 
         # Now advance time by 20 hours (above 2^4 because we made 4 attemps) and
         # see if we can now backfill it
         self.reactor.advance(datetime.timedelta(hours=20).total_seconds())
 
-        # Try again after we advanced enough time and we should see "b3" again
+        # Try at "A" again after we advanced enough time and we should see "b3" again
         backfill_points = self.get_success(
-            self.store.get_backfill_points_in_room(room_id)
+            self.store.get_backfill_points_in_room(room_id, depth_map["A"], limit=100)
         )
         backfill_event_ids = [backfill_point[0] for backfill_point in backfill_points]
-        self.assertListEqual(
-            backfill_event_ids, ["b6", "b5", "b4", "2", "b3", "b2", "b1"]
+        self.assertEqual(backfill_event_ids, ["b3", "b2", "b1"])
+
+    def test_get_backfill_points_in_room_works_after_many_failed_pull_attempts_that_could_naively_overflow(
+        self,
+    ) -> None:
+        """
+        A test that reproduces #13929 (Postgres only).
+
+        Test to make sure we can still get backfill points after many failed pull
+        attempts that cause us to backoff to the limit. Even if the backoff formula
+        would tell us to wait for more seconds than can be expressed in a 32 bit
+        signed int.
+        """
+        setup_info = self._setup_room_for_backfill_tests()
+        room_id = setup_info.room_id
+        depth_map = setup_info.depth_map
+
+        # Pretend that we have tried and failed 10 times to backfill event b1.
+        for _ in range(10):
+            self.get_success(
+                self.store.record_event_failed_pull_attempt(room_id, "b1", "fake cause")
+            )
+
+        # If the backoff periods grow without limit:
+        # After the first failed attempt, we would have backed off for 1 << 1 = 2 hours.
+        # After the second failed attempt we would have backed off for 1 << 2 = 4 hours,
+        # so after the 10th failed attempt we should backoff for 1 << 10 == 1024 hours.
+        # Wait 1100 hours just so we have a nice round number.
+        self.reactor.advance(datetime.timedelta(hours=1100).total_seconds())
+
+        # 1024 hours in milliseconds is 1024 * 3600000, which exceeds the largest 32 bit
+        # signed integer. The bug we're reproducing is that this overflow causes an
+        # error in postgres preventing us from fetching a set of backwards extremities
+        # to retry fetching.
+        backfill_points = self.get_success(
+            self.store.get_backfill_points_in_room(room_id, depth_map["A"], limit=100)
         )
 
+        # We should aim to fetch all backoff points: b1's latest backoff period has
+        # expired, and we haven't tried the rest.
+        backfill_event_ids = [backfill_point[0] for backfill_point in backfill_points]
+        self.assertEqual(backfill_event_ids, ["b3", "b2", "b1"])
+
     def _setup_room_for_insertion_backfill_tests(self) -> _BackfillSetupInfo:
         """
         Sets up a room with various insertion event backward extremities to test
@@ -938,18 +990,32 @@ class EventFederationWorkerStoreTestCase(tests.unittest.HomeserverTestCase):
 
     def test_get_insertion_event_backward_extremities_in_room(self):
         """
-        Test to make sure insertion event backward extremities are returned.
+        Test to make sure only insertion event backward extremities that are
+        older and come before the `current_depth` are returned.
         """
         setup_info = self._setup_room_for_insertion_backfill_tests()
         room_id = setup_info.room_id
+        depth_map = setup_info.depth_map
 
+        # Try at "insertion_eventB"
         backfill_points = self.get_success(
-            self.store.get_insertion_event_backward_extremities_in_room(room_id)
+            self.store.get_insertion_event_backward_extremities_in_room(
+                room_id, depth_map["insertion_eventB"], limit=100
+            )
         )
         backfill_event_ids = [backfill_point[0] for backfill_point in backfill_points]
-        self.assertListEqual(
-            backfill_event_ids, ["insertion_eventB", "insertion_eventA"]
+        self.assertEqual(backfill_event_ids, ["insertion_eventB", "insertion_eventA"])
+
+        # Try at "insertion_eventA"
+        backfill_points = self.get_success(
+            self.store.get_insertion_event_backward_extremities_in_room(
+                room_id, depth_map["insertion_eventA"], limit=100
+            )
         )
+        backfill_event_ids = [backfill_point[0] for backfill_point in backfill_points]
+        # Event "2" has a depth of 2 but is not included here because we only
+        # know the approximate depth of 5 from our event "3".
+        self.assertListEqual(backfill_event_ids, ["insertion_eventA"])
 
     def test_get_insertion_event_backward_extremities_in_room_excludes_events_we_have_attempted(
         self,
@@ -961,6 +1027,7 @@ class EventFederationWorkerStoreTestCase(tests.unittest.HomeserverTestCase):
         """
         setup_info = self._setup_room_for_insertion_backfill_tests()
         room_id = setup_info.room_id
+        depth_map = setup_info.depth_map
 
         # Record some attempts to backfill these events which will make
         # `get_insertion_event_backward_extremities_in_room` exclude them
@@ -973,12 +1040,15 @@ class EventFederationWorkerStoreTestCase(tests.unittest.HomeserverTestCase):
 
         # No time has passed since we attempted to backfill ^
 
+        # Try at "insertion_eventB"
         backfill_points = self.get_success(
-            self.store.get_insertion_event_backward_extremities_in_room(room_id)
+            self.store.get_insertion_event_backward_extremities_in_room(
+                room_id, depth_map["insertion_eventB"], limit=100
+            )
         )
         backfill_event_ids = [backfill_point[0] for backfill_point in backfill_points]
         # Only the backfill points that we didn't record earlier exist here.
-        self.assertListEqual(backfill_event_ids, ["insertion_eventB"])
+        self.assertEqual(backfill_event_ids, ["insertion_eventB"])
 
     def test_get_insertion_event_backward_extremities_in_room_attempted_event_retry_after_backoff_duration(
         self,
@@ -991,6 +1061,7 @@ class EventFederationWorkerStoreTestCase(tests.unittest.HomeserverTestCase):
         """
         setup_info = self._setup_room_for_insertion_backfill_tests()
         room_id = setup_info.room_id
+        depth_map = setup_info.depth_map
 
         # Record some attempts to backfill these events which will make
         # `get_backfill_points_in_room` exclude them because we
@@ -1027,13 +1098,15 @@ class EventFederationWorkerStoreTestCase(tests.unittest.HomeserverTestCase):
         # because we haven't waited long enough for this many attempts.
         self.reactor.advance(datetime.timedelta(hours=2).total_seconds())
 
-        # Make sure that "insertion_eventA" is not in the list because we've
-        # already attempted many times
+        # Try at "insertion_eventA" and make sure that "insertion_eventA" is not
+        # in the list because we've already attempted many times
         backfill_points = self.get_success(
-            self.store.get_insertion_event_backward_extremities_in_room(room_id)
+            self.store.get_insertion_event_backward_extremities_in_room(
+                room_id, depth_map["insertion_eventA"], limit=100
+            )
         )
         backfill_event_ids = [backfill_point[0] for backfill_point in backfill_points]
-        self.assertListEqual(backfill_event_ids, ["insertion_eventB"])
+        self.assertEqual(backfill_event_ids, [])
 
         # Now advance time by 20 hours (above 2^4 because we made 4 attemps) and
         # see if we can now backfill it
@@ -1042,12 +1115,12 @@ class EventFederationWorkerStoreTestCase(tests.unittest.HomeserverTestCase):
         # Try at "insertion_eventA" again after we advanced enough time and we
         # should see "insertion_eventA" again
         backfill_points = self.get_success(
-            self.store.get_insertion_event_backward_extremities_in_room(room_id)
+            self.store.get_insertion_event_backward_extremities_in_room(
+                room_id, depth_map["insertion_eventA"], limit=100
+            )
         )
         backfill_event_ids = [backfill_point[0] for backfill_point in backfill_points]
-        self.assertListEqual(
-            backfill_event_ids, ["insertion_eventB", "insertion_eventA"]
-        )
+        self.assertEqual(backfill_event_ids, ["insertion_eventA"])
 
 
 @attr.s
diff --git a/tests/storage/test_monthly_active_users.py b/tests/storage/test_monthly_active_users.py
index e8b4a5644b..c55c4db970 100644
--- a/tests/storage/test_monthly_active_users.py
+++ b/tests/storage/test_monthly_active_users.py
@@ -96,8 +96,12 @@ class MonthlyActiveUsersTestCase(unittest.HomeserverTestCase):
 
         # Test each of the registered users is marked as active
         timestamp = self.get_success(self.store.user_last_seen_monthly_active(user1))
+        # Mypy notes that one shouldn't compare Optional[int] to 0 with assertGreater.
+        # Check that timestamp really is an int.
+        assert timestamp is not None
         self.assertGreater(timestamp, 0)
         timestamp = self.get_success(self.store.user_last_seen_monthly_active(user2))
+        assert timestamp is not None
         self.assertGreater(timestamp, 0)
 
         # Test that users with reserved 3pids are not removed from the MAU table
@@ -166,10 +170,11 @@ class MonthlyActiveUsersTestCase(unittest.HomeserverTestCase):
         self.get_success(self.store.upsert_monthly_active_user(user_id2))
 
         result = self.get_success(self.store.user_last_seen_monthly_active(user_id1))
+        assert result is not None
         self.assertGreater(result, 0)
 
         result = self.get_success(self.store.user_last_seen_monthly_active(user_id3))
-        self.assertNotEqual(result, 0)
+        self.assertIsNone(result)
 
     @override_config({"max_mau_value": 5})
     def test_reap_monthly_active_users(self):
diff --git a/tests/storage/test_receipts.py b/tests/storage/test_receipts.py
index 9459ee1705..81253d0361 100644
--- a/tests/storage/test_receipts.py
+++ b/tests/storage/test_receipts.py
@@ -12,6 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+from typing import Collection, Optional
 
 from synapse.api.constants import ReceiptTypes
 from synapse.types import UserID, create_requester
@@ -84,6 +85,33 @@ class ReceiptTestCase(HomeserverTestCase):
             )
         )
 
+    def get_last_unthreaded_receipt(
+        self, receipt_types: Collection[str], room_id: Optional[str] = None
+    ) -> Optional[str]:
+        """
+        Fetch the event ID for the latest unthreaded receipt in the test room for the test user.
+
+        Args:
+            receipt_types: The receipt types to fetch.
+
+        Returns:
+            The latest receipt, if one exists.
+        """
+        result = self.get_success(
+            self.store.db_pool.runInteraction(
+                "get_last_receipt_event_id_for_user",
+                self.store.get_last_unthreaded_receipt_for_user_txn,
+                OUR_USER_ID,
+                room_id or self.room_id1,
+                receipt_types,
+            )
+        )
+        if not result:
+            return None
+
+        event_id, _ = result
+        return event_id
+
     def test_return_empty_with_no_data(self) -> None:
         res = self.get_success(
             self.store.get_receipts_for_user(
@@ -107,16 +135,10 @@ class ReceiptTestCase(HomeserverTestCase):
         )
         self.assertEqual(res, {})
 
-        res = self.get_success(
-            self.store.get_last_receipt_event_id_for_user(
-                OUR_USER_ID,
-                self.room_id1,
-                [
-                    ReceiptTypes.READ,
-                    ReceiptTypes.READ_PRIVATE,
-                ],
-            )
+        res = self.get_last_unthreaded_receipt(
+            [ReceiptTypes.READ, ReceiptTypes.READ_PRIVATE]
         )
+
         self.assertEqual(res, None)
 
     def test_get_receipts_for_user(self) -> None:
@@ -228,29 +250,17 @@ class ReceiptTestCase(HomeserverTestCase):
         )
 
         # Test we get the latest event when we want both private and public receipts
-        res = self.get_success(
-            self.store.get_last_receipt_event_id_for_user(
-                OUR_USER_ID,
-                self.room_id1,
-                [ReceiptTypes.READ, ReceiptTypes.READ_PRIVATE],
-            )
+        res = self.get_last_unthreaded_receipt(
+            [ReceiptTypes.READ, ReceiptTypes.READ_PRIVATE]
         )
         self.assertEqual(res, event1_2_id)
 
         # Test we get the older event when we want only public receipt
-        res = self.get_success(
-            self.store.get_last_receipt_event_id_for_user(
-                OUR_USER_ID, self.room_id1, [ReceiptTypes.READ]
-            )
-        )
+        res = self.get_last_unthreaded_receipt([ReceiptTypes.READ])
         self.assertEqual(res, event1_1_id)
 
         # Test we get the latest event when we want only the private receipt
-        res = self.get_success(
-            self.store.get_last_receipt_event_id_for_user(
-                OUR_USER_ID, self.room_id1, [ReceiptTypes.READ_PRIVATE]
-            )
-        )
+        res = self.get_last_unthreaded_receipt([ReceiptTypes.READ_PRIVATE])
         self.assertEqual(res, event1_2_id)
 
         # Test receipt updating
@@ -259,11 +269,7 @@ class ReceiptTestCase(HomeserverTestCase):
                 self.room_id1, ReceiptTypes.READ, OUR_USER_ID, [event1_2_id], None, {}
             )
         )
-        res = self.get_success(
-            self.store.get_last_receipt_event_id_for_user(
-                OUR_USER_ID, self.room_id1, [ReceiptTypes.READ]
-            )
-        )
+        res = self.get_last_unthreaded_receipt([ReceiptTypes.READ])
         self.assertEqual(res, event1_2_id)
 
         # Send some events into the second room
@@ -282,11 +288,7 @@ class ReceiptTestCase(HomeserverTestCase):
                 {},
             )
         )
-        res = self.get_success(
-            self.store.get_last_receipt_event_id_for_user(
-                OUR_USER_ID,
-                self.room_id2,
-                [ReceiptTypes.READ, ReceiptTypes.READ_PRIVATE],
-            )
+        res = self.get_last_unthreaded_receipt(
+            [ReceiptTypes.READ, ReceiptTypes.READ_PRIVATE], room_id=self.room_id2
         )
         self.assertEqual(res, event2_1_id)
diff --git a/tests/storage/test_registration.py b/tests/storage/test_registration.py
index 853a93afab..05ea802008 100644
--- a/tests/storage/test_registration.py
+++ b/tests/storage/test_registration.py
@@ -16,9 +16,10 @@ from twisted.test.proto_helpers import MemoryReactor
 from synapse.api.constants import UserTypes
 from synapse.api.errors import ThreepidValidationError
 from synapse.server import HomeServer
+from synapse.types import JsonDict, UserID
 from synapse.util import Clock
 
-from tests.unittest import HomeserverTestCase
+from tests.unittest import HomeserverTestCase, override_config
 
 
 class RegistrationStoreTestCase(HomeserverTestCase):
@@ -48,6 +49,7 @@ class RegistrationStoreTestCase(HomeserverTestCase):
                 "user_type": None,
                 "deactivated": 0,
                 "shadow_banned": 0,
+                "approved": 1,
             },
             (self.get_success(self.store.get_user_by_id(self.user_id))),
         )
@@ -166,3 +168,101 @@ class RegistrationStoreTestCase(HomeserverTestCase):
             ThreepidValidationError,
         )
         self.assertEqual(e.value.msg, "Validation token not found or has expired", e)
+
+
+class ApprovalRequiredRegistrationTestCase(HomeserverTestCase):
+    def default_config(self) -> JsonDict:
+        config = super().default_config()
+
+        # If there's already some config for this feature in the default config, it
+        # means we're overriding it with @override_config. In this case we don't want
+        # to do anything more with it.
+        msc3866_config = config.get("experimental_features", {}).get("msc3866")
+        if msc3866_config is not None:
+            return config
+
+        # Require approval for all new accounts.
+        config["experimental_features"] = {
+            "msc3866": {
+                "enabled": True,
+                "require_approval_for_new_accounts": True,
+            }
+        }
+        return config
+
+    def prepare(self, reactor: MemoryReactor, clock: Clock, hs: HomeServer) -> None:
+        self.store = hs.get_datastores().main
+        self.user_id = "@my-user:test"
+        self.pwhash = "{xx1}123456789"
+
+    @override_config(
+        {
+            "experimental_features": {
+                "msc3866": {
+                    "enabled": True,
+                    "require_approval_for_new_accounts": False,
+                }
+            }
+        }
+    )
+    def test_approval_not_required(self) -> None:
+        """Tests that if we don't require approval for new accounts, newly created
+        accounts are automatically marked as approved.
+        """
+        self.get_success(self.store.register_user(self.user_id, self.pwhash))
+
+        user = self.get_success(self.store.get_user_by_id(self.user_id))
+        assert user is not None
+        self.assertTrue(user["approved"])
+
+        approved = self.get_success(self.store.is_user_approved(self.user_id))
+        self.assertTrue(approved)
+
+    def test_approval_required(self) -> None:
+        """Tests that if we require approval for new accounts, newly created accounts
+        are not automatically marked as approved.
+        """
+        self.get_success(self.store.register_user(self.user_id, self.pwhash))
+
+        user = self.get_success(self.store.get_user_by_id(self.user_id))
+        assert user is not None
+        self.assertFalse(user["approved"])
+
+        approved = self.get_success(self.store.is_user_approved(self.user_id))
+        self.assertFalse(approved)
+
+    def test_override(self) -> None:
+        """Tests that if we require approval for new accounts, but we explicitly say the
+        new user should be considered approved, they're marked as approved.
+        """
+        self.get_success(
+            self.store.register_user(
+                self.user_id,
+                self.pwhash,
+                approved=True,
+            )
+        )
+
+        user = self.get_success(self.store.get_user_by_id(self.user_id))
+        self.assertIsNotNone(user)
+        assert user is not None
+        self.assertEqual(user["approved"], 1)
+
+        approved = self.get_success(self.store.is_user_approved(self.user_id))
+        self.assertTrue(approved)
+
+    def test_approve_user(self) -> None:
+        """Tests that approving the user updates their approval status."""
+        self.get_success(self.store.register_user(self.user_id, self.pwhash))
+
+        approved = self.get_success(self.store.is_user_approved(self.user_id))
+        self.assertFalse(approved)
+
+        self.get_success(
+            self.store.update_user_approval_status(
+                UserID.from_string(self.user_id), True
+            )
+        )
+
+        approved = self.get_success(self.store.is_user_approved(self.user_id))
+        self.assertTrue(approved)
diff --git a/tests/unittest.py b/tests/unittest.py
index 00cb023198..5116be338e 100644
--- a/tests/unittest.py
+++ b/tests/unittest.py
@@ -734,7 +734,9 @@ class HomeserverTestCase(TestCase):
             event.internal_metadata.soft_failed = True
 
         self.get_success(
-            event_creator.handle_new_client_event(requester, event, context)
+            event_creator.handle_new_client_event(
+                requester, events_and_context=[(event, context)]
+            )
         )
 
         return event.event_id
diff --git a/tests/util/caches/test_descriptors.py b/tests/util/caches/test_descriptors.py
index 48e616ac74..90861fe522 100644
--- a/tests/util/caches/test_descriptors.py
+++ b/tests/util/caches/test_descriptors.py
@@ -13,7 +13,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 import logging
-from typing import Set
+from typing import Iterable, Set, Tuple
 from unittest import mock
 
 from twisted.internet import defer, reactor
@@ -1008,3 +1008,34 @@ class CachedListDescriptorTestCase(unittest.TestCase):
             obj.inner_context_was_finished, "Tried to restart a finished logcontext"
         )
         self.assertEqual(current_context(), SENTINEL_CONTEXT)
+
+    def test_num_args_mismatch(self):
+        """
+        Make sure someone does not accidentally use @cachedList on a method with
+        a mismatch in the number args to the underlying single cache method.
+        """
+
+        class Cls:
+            @descriptors.cached(tree=True)
+            def fn(self, room_id, event_id):
+                pass
+
+            # This is wrong ❌. `@cachedList` expects to be given the same number
+            # of arguments as the underlying cached function, just with one of
+            # the arguments being an iterable
+            @descriptors.cachedList(cached_method_name="fn", list_name="keys")
+            def list_fn(self, keys: Iterable[Tuple[str, str]]):
+                pass
+
+            # Corrected syntax ✅
+            #
+            # @cachedList(cached_method_name="fn", list_name="event_ids")
+            # async def list_fn(
+            #     self, room_id: str, event_ids: Collection[str],
+            # )
+
+        obj = Cls()
+
+        # Make sure this raises an error about the arg mismatch
+        with self.assertRaises(Exception):
+            obj.list_fn([("foo", "bar")])
diff --git a/tests/util/test_check_dependencies.py b/tests/util/test_check_dependencies.py
index 5d1aa025d1..6913de24b9 100644
--- a/tests/util/test_check_dependencies.py
+++ b/tests/util/test_check_dependencies.py
@@ -40,7 +40,10 @@ class TestDependencyChecker(TestCase):
     def mock_installed_package(
         self, distribution: Optional[DummyDistribution]
     ) -> Generator[None, None, None]:
-        """Pretend that looking up any distribution yields the given `distribution`."""
+        """Pretend that looking up any package yields the given `distribution`.
+
+        If `distribution = None`, we pretend that the package is not installed.
+        """
 
         def mock_distribution(name: str):
             if distribution is None:
@@ -81,7 +84,7 @@ class TestDependencyChecker(TestCase):
                 self.assertRaises(DependencyException, check_requirements)
 
     def test_checks_ignore_dev_dependencies(self) -> None:
-        """Bot generic and per-extra checks should ignore dev dependencies."""
+        """Both generic and per-extra checks should ignore dev dependencies."""
         with patch(
             "synapse.util.check_dependencies.metadata.requires",
             return_value=["dummypkg >= 1; extra == 'mypy'"],
@@ -142,3 +145,16 @@ class TestDependencyChecker(TestCase):
             with self.mock_installed_package(new_release_candidate):
                 # should not raise
                 check_requirements()
+
+    def test_setuptools_rust_ignored(self) -> None:
+        """Test a workaround for a `poetry build` problem. Reproduces #13926."""
+        with patch(
+            "synapse.util.check_dependencies.metadata.requires",
+            return_value=["setuptools_rust >= 1.3"],
+        ):
+            with self.mock_installed_package(None):
+                # should not raise, even if setuptools_rust is not installed
+                check_requirements()
+            with self.mock_installed_package(old):
+                # We also ignore old versions of setuptools_rust
+                check_requirements()
diff --git a/tests/utils.py b/tests/utils.py
index 65db437697..045a8b5fa7 100644
--- a/tests/utils.py
+++ b/tests/utils.py
@@ -270,9 +270,7 @@ class MockClock:
         *args: P.args,
         **kwargs: P.kwargs,
     ) -> None:
-        # This type-ignore should be redundant once we use a mypy release with
-        # https://github.com/python/mypy/pull/12668.
-        self.loopers.append(Looper(function, interval / 1000.0, self.now, args, kwargs))  # type: ignore[arg-type]
+        self.loopers.append(Looper(function, interval / 1000.0, self.now, args, kwargs))
 
     def cancel_call_later(self, timer: Timer, ignore_errs: bool = False) -> None:
         if timer.expired: