When you’re hunting for bugs with a nondeterministic test, how many times should you run the test? The program might be deterministic: the same input produces the same behavior. Presuming you’re confident this is the case, you could fuzz the input for a while, hoping to catch bugs. But if your program is nondeterministic, you must not only give it random inputs, you must re-test it on the same input many times, hoping to achieve high branch coverage and catch rare bugs. I was at the first...
