I read this book once called Crisis Management. The main idea was to minimize the consequences of rare actions. So, even though train cars carrying lethal chemicals rarely derail, you build the train tracks so if one were to derail, say on a sharp curve, it derails far from populated areas. Or, even though nuclear reactors rarely release deadly radiocative gases, you don't allow people to build shanties nearby, where they would get lethal doses of radiation if there were an accident.
You can't get rid of adverse cases completely in most real-world situations. But you can work to minimize the consequences when the adverse case shows up. There are times when improving the worst case is pretty cheap, if you think ahead.
Policing is like this. The police very rarely need to use force ('rarely' in the sense of force used/interaction with civilian.) But when they do it makes sense to reduce the worst case: that the force will cause mortal injury. In the MOVE case, just wait them out. There is definitely increased cost, but it improves the expected value of the worst case by much more than the cost. Or, as I saw someone ask, speculatively I have to assume, on Twitter 'what would you do if some 6'4" guy was running at you threateningly?' The rational answer might be 'close your car window and call for backup.' It's never 'get out of the car and shoot him in the head.'