I'm at one of the latter, so I feel this in my bones. I've watched what should have been an innocent config change snowball into a pair of VM clusters shitting back and forth for 2 hours. Implemented strict change control that day. Kind of a pain, but the team learned a lot that day!