Handle poller resubmission during shutdown

[frojasg]

What changed?

Catch RejectedExecutionException when a poll loop re-submits itself and its executor has shut down. Rejections from an executor that is still running continue to reach the uncaught exception handler.

Why?

MultiThreadedPoller checks shouldTerminate() before it schedules the next poll. Shutdown can start between that check and execute(this). The resulting rejection escapes from finally, bypasses the shutdown-aware handler, and produces an ERROR log during normal worker shutdown.

The catch handles that race without changing task handoff or shutdown ordering. It checks the executor's shutdown state, so thread interruption alone cannot suppress an unexpected rejection.

Validation

The two new shutdown-race tests fail on the original code and pass with this change. Tests also verify that running-executor rejections remain visible, including on an interrupted thread.

On Java 21, formatting, SDK assembly, and all 13 tests across MultiThreadedPollerTest, GracefulPollShutdownTest, ShutdownManagerTest, and WorkerShutdownTest passed. The full test suite was not run locally.

Breaking changes?

None. No public API changes.

Server PR

None required.

Generated with Codex

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}