[reality_czech]

I've seen race conditions that only happen 1 time in 1000. Or race conditions that only happen when the machine is under load. I agree that it's frustrating to have to debug non-deterministic code, but I feel like there must be a better way to compare test runs than memcmp on the resutls.

You are right that deterministic pseudo-random numbers can be used when "random" inputs are required. I overlooked that... thanks for pointing it out.

[varelse]

One way I addressed this exact situation: Compute a hash on the state after each iteration and save it. Now run again normally and save several iterations worth of state somewhere. When the hash diverges, you have hit the race condition.

Do this a few times until you've characterized what's happening and your brain figures out how and why by seeing the pattern this will (hopefully) reveal.