Well, okay, that's a good point but I'm not crazy enough to want to deal with fork and Windows (use threads;)

My point is that most UNIX like OSes treat a return of zero from waitpid with the WNOHANG option as a special case that says "there are no stopped, continued or exited children". If you idle loop waiting for your children to exit, you won't be getting out of the signal handler often enough to handle new connections. Just testing for -1 isn't enough.