If you run a simulation using a pesudo-random generator, you may expect your measurements (of some quantity, and observable for instance) to be distributed according to some statistics. If your results do not satisfy these statistics, then you should pick another generator. If you know a good random-number generator (you usually try to avoid this kind because they are time-consumming), it is a good idea to compare your results for a short run, to see is they agree or not.