OP describes systems which seed tests with non-determinism to catch users who write tests relying on undefined behavior, such as order of equal elements. Writing tests against these systems is a game changer (once you figure out what’s going on and why there are flakes). I’m a huge fan of this subtle change which has spared me at least one bug. It also allows library authors to gain the advantage of not accidentally producing a behavior that induces Hyrum’s law. Kudos to the team for making this work at scale.