You are right on many points here, but I think you are completely wrong on randomness point. Randomness from the model, randomness from real word being non-dete

You are right on many points here, but I think you are completely wrong on randomness point. Randomness from the model, randomness from real word being non-deterministic and randomness from bugs in the code are qualitatively different. Yes, you can't make real-world experiment to produce exactly the same values all the time (if you did, you'd probably suspect something is wrong...). Yes, model can be probabilistic and return different results. But if different results come from race conditions in the code, and these are significantly different, then the model doesn't model what it is supposed to model, it's no longer the model of the real world. It's a model or the real world plus buggy code with random CPU scheduling. It's fine if we lived in a computer simulation running on a buggy VM with race conditions - then it would model our would perfectly. But since there's no indications we are in such world, the error introduced by this bug - and it looks it's not negligible at all - makes model diverge from the real world, and decisions regarding the real world made based on this wrong model are also wrong. It's a big problem, and no amount of other deficiencies of the article - which you appropriately highlighted - diminishes this very serious problem.