[Bug] Python 3.13.3 on macOS Intel hangs post-test

Start by reproducing the post-pytest hang in CI with Python 3.13.3 on Intel macOS, focusing on the worker shutdown path and the gRPC retry warnings. Trace which test or worker remains active after pytest finishes; done means the test run exits cleanly without the hang or repeated connection-refused retries.

Describe the bug

On latest Python (3.13.3 as of this writing), macOS Intel sometimes hangs post pytest run in our CI, with things like:

 2025-04-17T16:21:46.354310Z  WARN temporal_client::retry: gRPC call poll_nexus_task_queue retried 6 times error=Status { code: Unavailable, message: "tcp connect error: Connection refused (os error 61)", source: Some(tonic::transport::Error(Transport, ConnectError(ConnectError("tcp connect error", Os { code: 61, kind: ConnectionRefused, message: "Connection refused" })))) }
2025-04-17T16:21:46.373735Z  WARN temporal_client::retry: gRPC call poll_workflow_task_queue retried 6 times error=Status { code: Unavailable, message: "tcp connect error: Connection refused (os error 61)", source: Some(tonic::transport::Error(Transport, ConnectError(ConnectError("tcp connect error", Os { code: 61, kind: ConnectionRefused, message: "Connection refused" })))) }

and

2025-04-17T22:02:54.232127Z  WARN temporal_client::retry: gRPC call get_workflow_execution_history retried 5 times error=Status { code: Unavailable, message: "tcp connect error: Connection refused (os error 61)", source: Some(tonic::transport::Error(Transport, ConnectError(ConnectError("tcp connect error", Os { code: 61, kind: ConnectionRefused, message: "Connection refused" })))) }
2025-04-17T22:02:55.213783Z  WARN temporal_client::retry: gRPC call get_workflow_execution_history retried 6 times error=Status { code: Unavailable, message: "tcp connect error: Connection refused (os error 61)", source: Some(tonic::transport::Error(Transport, ConnectError(ConnectError("tcp connect error", Os { code: 61, kind: ConnectionRefused, message: "Connection refused" })))) }

This may be some worker in some test unable to shutdown properly in this specific situation.