> Or there are "reverts and code reviews are faster"

This seems like a slight misreading of the comment you're responding to. The claim is not that reverts are "faster", whatever that would mean; the claim is that the revert rate is lower.

Also, how would skimping on reviews lead to a lower revert rate? If anything, I'd imagine the normal assumption would be precisely the opposite - that skimping on reviews should lead to a higher revert rate, which is contrary to what the Android team found.