"Non-deterministic tests" usually refers to a test whose output depends on the execution environment in a way that cannot be controlled. For example, a multithr

"Non-deterministic tests" usually refers to a test whose output depends on the execution environment in a way that cannot be controlled. For example, a multithreaded test with a race condition, or a test that uses the real-time clock. This is fundamentally different from a test that uses random numbers. Another way to look at it: Deterministic tests can be used in `git blame`. A unit test that uses a specific PRNG algorithm and sets the seed is fine. A test with a race condition is not. Fuzzing is very useful, but it's not unit testing. If you want to run a fuzzer or some other kind of endless randomized testing in CI, it should be a separate job from the unit tests (IMO).