Issue type

Bug

Have you reproduced the bug with TensorFlow Nightly?

Yes

Source

source

TensorFlow version

2.20.0rc0

Custom code

Yes

OS platform and distribution

MacOS ARM

Mobile device

No response

Python version

Any

Bazel version

No response

GCC/compiler version

No response

CUDA/cuDNN version

No response

GPU model and memory

No response

Current behavior?

When importing TF 2.20.0rc0 and PyArray, the second import fails with:

libc++abi: terminating due to uncaught exception of type std::__1::system_error: mutex lock failed: Invalid argument
/Users/runner/work/_temp/fd5e114c-4a23-4ae6-b602-e37faa3477b3.sh: line 2:  1581 Abort trap: 6

Standalone code to reproduce the issue

Here's a minimal repository reproducing this issue: https://github.com/mtsokol/pyarrow-tf-repro

Here's a GitHub Actions job: https://github.com/mtsokol/pyarrow-tf-repro/actions/runs/16829225796/job/47672680245

python -m pip install tensorflow==2.20.0rc0 pyarrow==21.0.0
python -c 'import tensorflow; import pyarrow'

Relevant log output