> Since it's possible that block timestamps aren't chronological in order, what would happen if a block following a size increase trigger is back in the past before the size increase? Would that block have a lower size restriction again? Would using block height not be a more stable number to work with?

In the nVersion bits proposal that I co-authored we solved that issue by
comparing the timestamp against the median time, which is guaranteed by
the protocol rules to monotonically advance.