[Shubham-Padkonde]

Calling QueuedLogHandler.flush() currently stops its background worker. Records emitted afterward remain queued instead of being sent. This addresses the flush/stop behavior reported in #203.

Make flush() wait for outstanding sends, bounded by close_wait, while leaving the worker available for subsequent records. Move worker shutdown to close(). Queue entries are marked complete after sending (including error handling), so flushing also waits for a batch that the worker has already dequeued. During shutdown, check that the queue is empty before exiting after an empty batch.

Add real-worker regression tests for repeated flushes, empty flushes, an in-flight send, the flush timeout, send failures, and close. The initial five tests failed before the change. The fix does not make a handler reusable after close() or logging reconfiguration that closes it.

Validation: python -m pytest tests/unit tests/ci/build-test -q — 63 passed, 8 skipped on Python 3.12. Live SLS integration and legacy Python runtimes were not exercised locally.

[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.}