Okay, restarted the process. Things are back to normal. Not sure why was the process taking up the entire CPU.
It’s possible that some query never dies, but never finishes either; so the servers are busy trying to process it over and over again. We should set up some generous timeouts for queries – just a safe thing to do.