Oops. You're right. In fact I tried that for the original code (using a very similar code as the above example shows in context of a Proc::Daemon fork) where I couldn't see any change. And since I could reproduce the very same behavior with the above code, I didn't do the check again for the example code. Thanks for that hint!

However I think I was going the wrong way anyways, I will recode the queuing system as ithreads.