We incorrectly didn't use the returned `Responder` if the client had disconnected, which meant that the resource used by the Responder wasn't correctly released. In particular, this exhausted the thread pools so that *all* requests timed out.
registration_shared_secret
synapse.api.auth.Auth
Requester
UserID
_check_sigs_and_hash_and_fetch
/make_knock
/forget
/messages
get_users_in_room
event_map
unpersisted_events
event_push_actions_staging
destination_is