I'm not sure I understand the wins of this over just using the standard multiprocessing module. You can trivially define a worker pool or arbitrary size and pi

I'm not sure I understand the wins of this over just using the standard multiprocessing module. You can trivially define a worker pool or arbitrary size and pipe a stream of values through your function using multiple cores, avoiding GIL issues. If you have something which scales to multiple machines just use celery or redis streams.