You are confusing two things here. Flaky or flappy tests are those tests that fail non-deterministically for the exactly same input. That is, it is impossible to reliably reproduce the behavior even when given the same input, and is usually caused by dependence on some environment factor. Randomly generated tests on the other hand are not flappy or flaky. One can indeed save the randomly generated input, and reproduce the same behavior if one observes a crash. Indeed, it is pretty easy to do that if you save the random seed in most property based testing systems.