grpc: Python `grpcio` 1.52.0 is broken on Apple Silicon

What version of gRPC and what language are you using?

grpcio 1.52.0, installed via PyPI (using Poetry) CPython 3.9.16

What operating system (Linux, Windows,…) and version?

macOS 13.1

What runtime / compiler are you using (e.g. python version or version of gcc)

CPython 3.9.16, installed via pyenv, native Apple Silicon on an M1 MacBook

What did you do?

Please provide either 1) A unit test for reproducing the bug or 2) Specific steps for us to follow to reproduce the bug. If there’s not enough information to debug the problem, gRPC team may close the issue at their discretion. You’re welcome to re-open the issue once you have a reproduction.

Install a development build of the flwr library (Flower, flower.dev) and run one of our code examples (e.g. quickstart_pytorch). This should not be necessary though, I assume that the issue is reproducible in the same environment in other ways, too.

What did you expect to see?

The example completes successfully.

What did you see instead?

When running the Flower server (via flower-server), we get a repeated log output:

E0206 11:46:24.733515000 12999757824 thread_pool.cc:254]               Waiting for thread pool to idle before forking

This only happens with grpcio 1.52.0, not with grpcio 1.51.1, which is why we decided to disallow 1.52.0 in our next release (https://github.com/adap/flower/pull/1644).

Update: this blocks the process and the terminal session, the only way to kill it is to use kill -9 in a different terminal window.

Anything else we should know about your project / environment?

Native Apple Silicon, no Rosetta2 x86 emulation.