On average, the D less R margin in the early vote mispredicted the final Clinton/Trump margin by 14 points! Pollsters get yelled at when their polls are off by even 3 points, and anything more than that is considered an absolute disaster. Imagine if a poll was off by 14 points: no one would ever listen to it again! And yet we get the same frankly amateurish analysis of the early vote in every election.
I think that’s sort of the point - if 2016 was our last “normal” election and early voting wasn’t prognosticative of election results then, there’s no hope it would be anything other than more variable and chaotic now.
The point wasn’t about a “return to normal” or else he would be saying it was an indicator.