Ensure fed-sender catchup does not block for full state (#15248)

* Reproduce bad scenario in test
* Avoid catchup optimisation for partial state rooms

1 files changed, 7 insertions, 2 deletions

diff --git a/synapse/federation/sender/per_destination_queue.py b/synapse/federation/sender/per_destination_queue.py

index 478187ce44..31c5c2b7de 100644
--- a/synapse/federation/sender/per_destination_queue.py
+++ b/synapse/federation/sender/per_destination_queue.py
@@ -497,8 +497,8 @@ class PerDestinationQueue:
             #
             # Note: `catchup_pdus` will have exactly one PDU per room.
             for pdu in catchup_pdus:
-                # The PDU from the DB will be the last PDU in the room from
-                # *this server* that wasn't sent to the remote. However, other
+                # The PDU from the DB will be the newest PDU in the room from
+                # *this server* that we tried---but were unable---to send to the remote.
                 # servers may have sent lots of events since then, and we want
                 # to try and tell the remote only about the *latest* events in
                 # the room. This is so that it doesn't get inundated by events
@@ -516,6 +516,11 @@ class PerDestinationQueue:
                     # If the event is in the extremities, then great! We can just
                     # use that without having to do further checks.
                     room_catchup_pdus = [pdu]
+                elif await self._store.is_partial_state_room(pdu.room_id):
+                    # We can't be sure which events the destination should
+                    # see using only partial state. Avoid doing so, and just retry
+                    # sending our the newest PDU the remote is missing from us.
+                    room_catchup_pdus = [pdu]
                 else:
                     # If not, fetch the extremities and figure out which we can
                     # send.