That's fine for regular Python, but this doesn't convince me for multithreaded Python. Most Python modules which are optimized for performance (numpy, pytorch,

That's fine for regular Python, but this doesn't convince me for multithreaded Python. Most Python modules which are optimized for performance (numpy, pytorch, pandas, and all others built on top of them) are already multithreaded and drop the GIL so you can parallelize your workload with the threading module. If someone really needs several threads of pure python being interpreted, something is afoul imo.