Gracefully stop the worker: prevent new polls, allow any in-flight activation to finish without cancellation, and wait for currently active jobs to drain (be acknowledged) up to waitUpToMs. If timeout is reached, falls back to hard stop logic (cancels activation if still pending).
Optionalopts: { checkIntervalMs?: number; waitUpToMs?: number }
Begin polling for jobs. Safe to call at any point: the request is buffered until the transport is ready, and redundant calls (including an explicit call on an
autoStartworker) are dropped rather than starting a second poll loop. Once the worker is stopped,start()is a no-op.