Daniel Standish

> So as you cherry-pick into smaller PRs and leaving complexity manageable... if all is green might still be good to merge the pieces and I assume to make it...

also for reference @jscheffl here's the PRs that came before on this one 😅 https://github.com/apache/airflow/pulls?q=is%3Apr+project%3Aapache%2F169+is%3Aclosed

> Agree `RPC server` is a better name. Also marking it as private (how?) is a good idea. Maybe we rename rpcServer section “_rpcServer” until release?

> Why not. It will also be disabled by default so no big deal and we still have the feature flag `AIRFLOW_ENABLE_AIP_44` set to False by default that prevents the...

> Let's prefix with `_`, or maybe `_aip44RpcServer`? I'm hesitant of a docs only change. Did it. Went with `_rpcServer`. I think that's definitely clear enough. When we make it...

looks like what we need to do is find all the tasks “in the scope” of the teardown _(and that is, those tasks that are in between the teardown and...

We were able to improve the behavior with these settings: ``` AIRFLOW__CELERY_BROKER_TRANSPORT_OPTIONS__SOCKET_TIMEOUT=30 AIRFLOW__CELERY_BROKER_TRANSPORT_OPTIONS__SOCKET_CONNECT_TIMEOUT=5 AIRFLOW__CELERY_BROKER_TRANSPORT_OPTIONS__SOCKET_KEEPALIVE=True AIRFLOW__CELERY_BROKER_TRANSPORT_OPTIONS__RETRY_ON_TIMEOUT=True ``` In our repro (which was to force kill the redis pod) the celery worker...

> @dstandish dude, it's passed about 30 minutes since after applying your solution. my celery worker was throwing `connection reset by peer.` errors and freezing very frequently, but now the...

@uranusjr just going to merge this for now; if you look and find issues, we can make changes.

> @tirkarthi @dstandish do we want to merge this PR for 2.9.2? @eladkal I think we cannot merge this anymore now that https://github.com/apache/airflow/pull/39336 is merged. At least not into main....