It could happen to Monte Carlo method (although rarely), that during the simulation, we may get run length that is larger than what we usually expected (For example, in the case we want to test the average running length of a process, while the mean length could be 370, perhaps an exception happened and the running length reaches 5000, although very unlikely, but it goes out of our recording limit).
In this case, how do we treat this outlier? Do we simply eliminate it? Or generate a new one to replace it? Or there’s other ways to solve this problem?