Because you're all developers, you probably want to know more about the problem. It looks like we have one process that refuses to stop, sometimes, and stay there indefinitely. And when there are too many processes it starts to malfunction. That's why a restart helps, but we are still investigating why they refuse to stop.
If you encounter new problems please tell me. Just make sure that the test pass sometimes and your code is correct :). I've upgraded the servers so that should be fine for a while.