Fix deadlocked engine in RPC

[bcherry]

There's an issue where if you're busy handling an incoming rpc request, additional engine events will not be processed. this causes a deadlock essentially if you have nested RPC, since the incoming acks won't be processed, but also probably gums up plenty of other features of livekit...

this solution is probably not correct, but it works and i included an example. need help @typester !

[ilo-nanpa]

it seems like you haven't added any nanpa changeset files to this PR.

if this pull request includes changes to code, make sure to add a changeset, by writing a file to .nanpa/<unique-name>.kdl:

minor type="added" "Introduce frobnication algorithm"

refer to the manpage for more information.

[typester]

Interesting. Will take a look!

[typester]

The change you made (d7c8cda) seems like a good fix to me 👍

[typester]

What caught my attention in a different context is the await here:

rust-sdks/livekit/src/room/mod.rs

Line 649 in fea7557

task.await;

If we await at this point, engine_events will be processed one by one. To me, it seems more natural not to await here, but what was the intention behind this design?

[bcherry]

@typester I'm not sure about that, maybe @theomonnom knows

[theomonnom]

yeah, it is expected for engine events to be processed one by one (keeping the order)