Hey team, I have had multiple issues where the agent has taken up to 20-30 seconds to answer the call causing some hangups but then if called back immediately the agent picks up within a few seconds.
This is causing a lot of frustration, could you please look into why this is happening and how to resolve it?
call_256bc86a993f729fc09e0ea543d (04-25, 23:40 UTC) — the call itself was accepted instantly, but the agent worker took ~23.3s to be assigned to it. Our prod logs show a LongJoinRoomWait event of 23,338ms for this call. This was part of a platform-wide dispatch-queue backlog that lasted from roughly 23:00 UTC on 04-25 through ~13:00 UTC on 04-26. Once the worker was assigned, joinRoom itself took only 119ms — so the dead air was queue-side, not network or LiveKit-room latency. The caller hung up at ~30s.
call_e63304ee576fdff31df7ae78ca8 (04-27, 08:47 UTC) — same shape (~17s gap between “call setup complete” and “agent worker started”), but the call completed successfully (full intake, 2m 25s, agent_hangup). We did not find a matching LongJoinRoomWait log for this exact hour, so we can’t confirm the same root cause with the same direct evidence — the symptom matches but it’s likely a smaller, isolated dispatch hiccup.
call_b06b38ca7d50886dda4e643eefc (04-20) — this one is unrelated. It’s a web call that errored before any worker handoff (the client side never completed the audio connection). No dispatch delay was seen here.