You open the logs at night. The deploy finished six hours ago, everything is green, you looked thirty minutes ago.
The fear is not about the server. Checking is a way to feel control for a minute where there is none: the service can fall over at any moment, and an open tab will not change that. You are not getting information, you are getting a sensation.
It sets in simply, and it works against you. Anxiety offers an action. The action brings relief for about ten minutes. The system records: it worked. Next time the threshold is lower and the interval shorter.
That is what holds it up — not that it helps, but that it helps quickly. Fast reinforcement always beats useful reinforcement, and that is not a question of character.
There is one test that separates a working check from a ritual one: does it change what you do. If a red log has you fixing things at three in the morning, the check is real, it is part of being on call. If you look and go to sleep, you were not checking. You were soothing.
The difference is not a moral one, it is about what can be fixed. A real check is fixed with monitoring and an alert: let the system wake you when there is a reason. A ritual check is not fixed by monitoring, because it is about there being no reason and the anxiety being there anyway.
Everyone gets anxious. The thing to watch is not the check itself but the interval: when it starts shortening on its own, it is no longer about the deploy.