This test comparing the performance of regexp-split between strings and bytes fails a lot on Windows. Does it make sense to declare this test as unreliable on Windows? https://github.com/racket/racket/blob/master/pkgs/racket-test-core/tests/racket/string.rktl#L343

Yes, that looks like a test that should be wrapped in (when (run-unreliable-tests? 'timing) ….).

Okay. For now, let me wrap it when the platform is Windows. I haven’t seen it failing for other OSes.