I think you're right. I have python threads and processes meshed together in production code, and it's not pretty to look at. Sometimes it would crash for no reason, like this morning, when a job got stuck because one of its child process refused to die properly, thus blocking the join.