https://www.academia.edu/11274021/Cipolla_-The_Basic_Laws_of...
There are actually plenty of adversarial situations in which it's beneficial to give yourself a handicap in order to win. If you're playing "chicken" (where you drive cars at each other and whoever turns away first loses) you can blindfold yourself and remove the steering wheel. If you're doing Mutually Assured Destruction, you can set up an automatic retaliation system which you are unable to pause or cancel. By removing some of your options, you can effectively signal to your opponents so that they are forced to respond to your strategy.
Similarly, it can be beneficial to act "irrationally", because if your opponents are rational actors, they won't even try the strategy against you. E.g. If I know you will "irrationally" come for revenge after the fact even to your own detriment despite the fact that it won't fix anything, I might not try something against you in the first place.
I have seen this pursued with gusto as a career development strategy on occasion and it works rather well unfortunately. The basic approach is to be so outlandish, unprofessional, and unreliable that capable and rational colleagues will pick up the slack, confident that the crazy person will be quickly gotten rid of due to their bad behavior, however that very rarely happens. The irrational actor ends up being promoted because they are the most visible person on the project as everyone else is in hiding in anticipation of a shit-storm.
It can have appeal to some. And there is the fact that the standard 100% professional corporate culture is extremely unappealing and in contrast to such behavior.
The canonical example for OP is the “mad man” strategy Reagoj supposedly followed.
[0] Other people more qualified than me to discuss the matter have argued variously that he is merely a fan — superficial familiarity but missing the point — or that he is reflective of the wider Rationality movement.
For the sake of discussion; One could watch new inmates like a hawk for an initial probationary period and enact disciplinary action should they reveal themselves as described.
In fact, taking the idea to its (awful) logical conclusion, randomly enacting disciplinary action in the initial period regardless of inmate behaviour would prove effective. "Beatings will continue until morale improves" I believe is the saying.
If there's a pattern, it will be identified and exploited. New inmates being under closer observation or duress is a pattern, easy to spot and easy to test.