The problem with threat is that it sometimes paralyses people ("fight or flight or freeze"). Also, if the consequences of X are unpleasant, it conditions people also to not think about X.

So I would not be surprised to see such motivational systems fail. Humans don't maximize their utility functions. They are composed of subsystems, selected by evolution to more work than fail on average, but some kinds of inputs can still mess them up.

Threatening people if they do something undesired is sometimes just as efficient as kicking your TV set if it does not work properly. Whether it works or not, it makes the punisher feel good, and it can be socially justified, so we continue doing it even in absence of results.