Ballmer's organization would. I've seen it.
In Ballmer's organization, groups deliberately used head count to hire low-performers, who would then be put on the chopping block in review cycles. The horse-trading games in review meetings were brutal. People wrote self-reviews, but they didn't mean anything because it was all decided ahead of time. It was all a big kabuki dance pretending that the company could measure people accurately, even across widely separated groups (an SDE II in the Xbox group was not the same as an SDE II in Windows, believe me).
Oh, and God help you if you get into one of these situations: Say you're a seasoned veteran, a senior engineer, and your boss got out of school three years ago and has no idea how to defend you in stack ranking. Or your manager is an over-the-hill engineer who doesn't want to manage but was forced into it, and doesn't even buy you equipment, much less give a crap about you in a stressful meeting he doesn't want to attend in the first place. Or you're reporting to a political climber, who will throw you under the bus just to be a "yes man" to someone else in political power.
One year I was a "ten percenter" and in peril of losing my job. Two years later I was getting promotions and on the partner track. Go figure. We can't pretend that reviews are even remotely accurate given this scale of noise.
I think that deciding pay scale / ranking and deciding who should be let go should be a different set of criteria, and probably a different review cycle. Not totally unlinked, but not utterly congruent, either. This would also handle the "high paid asshole" who is productive, but totally toxic and doing damage far in excess of his pay.