Stabilize test-racy-mutations.t on Windows

@marmoute - this shows some of the differences around the failure. I don't see a way to alter the code to avoid this, and now that I see these tests, I'm wondering why there are external differences around the repo files (especially when the manifest is edited). I do see some minor differences in output when I load the code up with debugging, but let's start with "is there a problem with the setup here?", before trying to figure out where the code diverges.

I'm also fine with just landing the first few commits that conditionalize the output and call it a day. I can't tell how big of a problem this actually is.