my 2 cents:
I suspect they're able to look at the broad strokes and see if there are big differences in representations between the rounds. For example, this year... you'd probably expect to see a larger than usual number of finance types in rd 2. My guess (and i've heard some confirmation of this) is that when this happens, they just waitlist more people in those groupings than usual to be able to compare apples to apples. But this year is particularly unique given the timing of the implosion of much of wall street.
At the end of the day, the schools have no incentive to fully view the two rounds as independent. If they have more choice in rd 2, it makes sense to take the more borderline cases from rd 1 and compare them with their peers in rd 2.
So how does it work with R1 vs. R2?
R2 applications are already in, but some AdComs (Stanford, HBS) have 2 more weeks to make R1 decisions. Did they give themselves more time (compared to Booth, Wharton, Kellogg) to evaluate R2 pool? Is it at all possible to scan R2 so quickly and make decision for R1 at the same time? Or maybe these two rounds are assessed independently?