That's what I commented about this particular situation in the second point, right down to the caniuse reference. The first part about a cutoff point was in response to "Regressions are regressions. You don't get to make excuses if you unilaterally changed something for someone" which is not true or the problem in this situation.