Stuck messages in the timeline

[ara4n]

Over the last few weeks i've had ~4 instances where a sent message gets stuck at the bottom of the timeline because we didn't successfully receive an echo down /sync. I rageshaked; @richvdh took a look with jaeger and says there may be two competing /sync requests which tried to receive the message, and so it may have fallen between the cracks. As far as I know, eleweb should never be running overlapping sync loops - so it sounds like something is going worryingly wrong here.

[ara4n]

this is also matrix-org/synapse#9424 (comment)

[turt2live]

I'm slightly worried this is a synapse-side threading issue btw. Mjolnir and such are seeing duplicate events down /sync, which sounds like some sort of corrupted aggregated state.

[t3chguy]

Duplicate of #17471

[ara4n]

I don't think this is a precise dup of the #17471, given @richvdh's investigation with jaeger showed two possible /sync reqs from eleweb which would have caught the event. Whereas #17471 says...

This is a bug when the server never remote-echoes the event back down /sync

which is not the case here.

[richvdh]

My money is currently on synapse not returning the event in either sync request, to be clear.

[dbkr]

I just got this again and it's worse now because the new 'sent' tick is just on the latest message, so i you send another message afterwards, the latter one can get through but then the two messages just look like they've been re-ordered, and the 'sent' tick is on the one that actually didn't send at all!

[nrdxp]

Is anyone aware of a workaround to resolve this. It's quite annoying 😓

[JokerGermany]

Have the problem in the matrix HQ chat, too and it's really really anoying
https://matrix.to/#/!OGEhHVWSdvArJzumhm:matrix.org/$aX_HOZknt_1BUOI1WqtHpXQEcLVFahKsISxUEOyz0fc?via=matrix.org&via=libera.chat&via=tchncs.de

[novocaine]

@ara4n thinks this is an E2EE client bug .. need to understand if that's tracked elsewhere

[Palid]

This issue also exists in the Android version of the app. @kittykat are you aware of it?

[Palid]

If I remember correctly it usually happened whenever I had some backlog to sync and I sent the message and closed the app/changed room focus (either or) before backlog did properly sync, which left me with this hanging message.

[BillCarsonFr]

New occurence on web for me https://github.com/matrix-org/element-web-rageshakes/issues/10049

[t3chguy]

Closing in favour of more generic element-hq/element-meta#1597