Something on MacOS causes stat(1) to regularly fail to identify pipes, for some reason. I suspect yet another filesystem-race in the test setup, but don't understand yet what's going on.

Just to hammer home that this is a real issue, here are some PRs where I complained about it:

• cp: fix flaky test test_cp_arg_interactive_update, document adjacent bug #6020 (comment)
• date: fix interaction of flags, fix issues around --set #6142 (comment)

… plus all the times it flaked for other people.

Could someone with a Mac please try various "sleeps" to reproduce the issue, ideally fix it?