Assume some model, and do calculations within that model to tell you what the probability of some event is under that model. What is just about impossible to c

Assume some model, and do calculations within that model to tell you what the probability of some event is under that model. What is just about impossible to calculate is what the probability is that the model captures only way for X event to occur. I find people sometimes confuse the reliability of calculations within the model as a defense of the model itself - usually unintentionally, but sometimes intentionally (with industries defending the safety of some practice...).