A way for ASes to poke synapse to retry the timeout queue (SYN-774)

[matrixbot]

If an AS is down for a while, synapse rightly starts backing up the queue, retrying less often. But you can get it into a state where the retry timeout is so slow that it can take many minutes after restarting your AS before messages start flowing again.

It would be nice for synapse to have a button that ASes can poke to say "Hi, I started up now", and synapse can then try flushing the queue again early.

(I'll have a go at writing this now)

(Imported from https://matrix.org/jira/browse/SYN-774)

(Reported by @leonerd)

[matrixbot]

Jira watchers: @leonerd

[matrixbot]

On further thought, it seems that any entirely nontrivial bridge will probably start talking to synapse as soon as it starts. Current examples are that the IRC bridge performs a full /initialSync and that slack and gitter bridges request the state events out of their currently-mapped rooms.

It may therefore be sufficient that synapse just bumps ASes up the retry queue if it sees inbound requests from the AS's auth token.

-- @leonerd

[MadLittleMods]

Closing in favor of #14240