[eersnington]

#5860 runs the sleep hooks, including connection disconnects, after in-flight work drains. If that work outlives the grace period, connections can still be attached when the grace deadline fires.

• At the grace deadline, the actor closes every connection that grace closes: all of them on destroy, and the non-hibernatable ones on sleep. It closes the transports directly, because hooks dispatched at the deadline are never awaited.
• emit_grace_events uses the same closes_on_grace filter, so the hibernation rule lives in one place.
• Adds a drain test: an action in flight when the runner drains still returns its result to the client.

// rivetkit-rust/packages/rivetkit-core/src/actor/task.rs, on_sleep_grace_deadline
self.ctx
	.disconnect_conns(|conn| closes_on_grace(grace.reason, conn))
	.await

---

This is part 6 of 6 in a stack:

•  6  fix(rivetkit-core): close attached connections at the sleep grace deadline #5871 👈
•  5  fix(rivetkit): run onSleep after in-flight work and refuse dispatch once it starts #5860
•  4  fix(rivetkit): refuse actions after onSleep with actor.restarting #5876
•  3  fix(rivetkit-core): close drained connections with ws.going_away #5867
•  2  fix(pegboard-gateway): send the client close without waiting on the envoy #5874
•  1  fix(rivetkit): retry connect after guard actor_ready_timeout #5866

[claude]

Review

Small, focused fix. Extracting closes_on_grace so emit_grace_events and the deadline path share the hibernation rule is a good change.

Code quality

• The logic looks correct. disconnect_conns goes through disconnect_transport_only, which closes transports without dispatching hooks. That matches the comment that hooks at the deadline are never awaited.
• The warn! follows the logging conventions. Per repo guidance, consider adding a likely cause to the message.
• disconnect_transport_only awaits each connection serially. A slow or hung transport close could use up the rest of the deadline budget before run_handle.abort() and final cleanup. Consider a bounded timeout or concurrent closes.
• The close reason (ws.going_away) depends on actor_aborted(). Please confirm the abort signal is already set at the grace deadline for both sleep and destroy.

Potential issues

• Please confirm onDisconnect cannot fire twice, once from the deadline path and again from final cleanup, for connections closed here. The existing "runs onDisconnect once" test only covers the drain-with-hang case.

Tests

• The new driver test covers the happy path, where work finishes before the deadline. Nothing exercises the deadline path itself, where work outlives grace and connections are closed directly. A core test under tests/ with a short grace period and a hanging action would cover it.
• The reformatting of output.split(...) in the existing test is unrelated churn.

Docs

• The sleep-sequence.md update is good.

Overall this looks good. Main requests: a core-level test for the deadline path, and a look at serial close latency.