Yup, don't ever rely on any random hash function, even if widely used, to be repeatable between processes. I've seen a deployment of a small caching system that

Yup, don't ever rely on any random hash function, even if widely used, to be repeatable between processes. I've seen a deployment of a small caching system that used the Python's built in hash function. The author of the code was wondering why does he get a ~zero hit rate (processes were short lived in that setup). Related: iterating over hash maps is a very common way to get nondeterministic output. I've removed a lot of inconsistency bugs by fixing the hash map iteration order. Usually by just using the sorted map, since the performance of a randomly chosen piece of code in a large system is almost irrelevant.