This is to avoid division by zero, I assume, but that is a separate issue that could occur with the existing code in the same situation, i.e. if DELAY(1000) isn't enough to make them tick up (or they exactly wrap around). If you want to make that change it should be a separate commit, though start_TSC is supposed to be catching this possibility, no? (Not sure exactly what the conditions are here for when each snippet runs...)

Hm, being read-only is a bit different. The "test whether they really work" is to make sure that they tick up. I think you just want to add the read-only part as a separate sentence afterwards?

Rogue space before the closing parenthesis, and no need to add the braces

presumably?

It's true the same possibility exists with the original code. Given what is done is start_TSC(), that can appear redundant, but really we should prepare for weird emulator behaviors, and this test costs basically nothing, so IMO it should be done (whether in a separate commit or here in passing).

Typo here (existing patch should not compile).

This is my mistake in transcribing. I thought after the second indent we use 4 spaces, but I think I see it's more complex than that.

One nit