This caused a curious case of forking a new worker any time something unrelated to the worker code (but originating in the same parent process) forked and exited.
There's still race conditions in the code (such as trying to maintain a worker pool and missing some process exits while checking on others) which should probably fixed in another diff.
This caused a curious case of forking a new worker any time something unrelated to the worker code (but originating in the same parent process) forked and exited.
There's still race conditions in the code (such as trying to maintain a worker pool and missing some process exits while checking on others) which should probably fixed in another diff.