Avoid re-computing the event ID when cloning events. (#19527)

`event_id` is a lazily-computed property on events, as it's a hash of
the event content on room version 3 and later. The reason we do this is
that it helps finding database inconsistencies by not trusting the event
ID we got from the database.

The thing is, when we clone events (to return them through /sync or
/messages for example) we don't copy the computed hash if we already
computed it, duplicating the work. This copies the internal `_event_id`
property.

Author: sandhose

changelog.d/19527.misc

@@ -0,0 +1 @@
+Avoid re-computing the event ID when cloning events.

synapse/events/utils.py

@@ -48,7 +48,7 @@
 from synapse.logging.opentracing import SynapseTags, set_tag, trace
 from synapse.types import JsonDict, Requester
 
-from . import EventBase, StrippedStateEvent, make_event_from_dict
+from . import EventBase, FrozenEventV2, StrippedStateEvent, make_event_from_dict
 
 if TYPE_CHECKING:
     from synapse.handlers.relations import BundledAggregations
@@ -109,6 +109,14 @@ def clone_event(event: EventBase) -> EventBase:
         event.get_dict(), event.room_version, event.internal_metadata.get_dict()
     )
 
+    # Starting FrozenEventV2, the event ID is an (expensive) hash of the event. This is
+    # lazily computed when we get the FrozenEventV2.event_id property, then cached in
+    # _event_id field. Later FrozenEvent formats all inherit from FrozenEventV2, so we
+    # can use the same logic here.
+    if isinstance(event, FrozenEventV2) and isinstance(new_event, FrozenEventV2):
+        # If we already pre-computed the event ID, use it.
+        new_event._event_id = event._event_id
+
     # Copy the bits of `internal_metadata` that aren't returned by `get_dict`.
     new_event.internal_metadata.stream_ordering = (
         event.internal_metadata.stream_ordering