Pretty sure with the new server implementation, it's easy to get a zombie task graph. It might happen every time the stop button is hit.
Pretty sure we need to handle cancel errors in TaskGraph's run, and cancel all the Tasks it started. Tasks might also need to handle cancel errors - it's unclear
Pretty sure with the new server implementation, it's easy to get a zombie task graph. It might happen every time the stop button is hit.
Pretty sure we need to handle cancel errors in TaskGraph's run, and cancel all the Tasks it started. Tasks might also need to handle cancel errors - it's unclear