_run_push_actions_and_persist_event re-starts a finished logging context when persist_events_and_notify raises an exception

[squahtx]

When persist_events_and_notify raises an exception, we try to clean up push actions in a background task:

synapse/synapse/handlers/federation_event.py

Lines 1961 to 1969 in e3163e2

	try:
	await self.persist_events_and_notify(
	event.room_id, [(event, context)], backfilled=backfilled
	)
	except Exception:
	run_in_background(
	self._store.remove_push_actions_from_staging, event.event_id
	)
	raise

run_in_background inherits the current logging context, but since we don't await the background task, we end up closing the logging context before the background task finishes. We ought to either await remove_push_actions_from_staging directly, or create a logging context for it.

[squahtx]

Actually every call to run_in_background where we discard the return value probably has the same issue.

[clokep]

Actually every call to run_in_background where we discard the return value probably has the same issue.

I think grep -r " run_in_background" synapse (note the two spaces before it) would be a quick way to find those. I see a bunch. 😢