You altered the history when you told the S9v1 to leave without a long enough sample size. Had they stayed the variance would of caught back up. Any sample size should be bare minimum one year.
He didn't "alter" anything. S9s will forever suffer from the pre-January deficit and would have to span months of overages to "break even"; this is not the same as "altering history", it's forcing historical losses to coincide with current statistics.
Except that, as I said, people have still been mining with S9v1 on the pool non-stop.
Just a lot fewer.