The assumption is that Y Combinator thinks that they're providing value through resources like mentorship, and that if they spread those resources too thin, they'll reduce the odds of producing a billion dollar company (or some similar metric). Sure, that limit may be flexible a bit one way or the other way based on the quality of the applicants, but it's neither necessarily irrational nor artificial prestige-bumping quota to limit to a couple of dozen teams rather than taking hundreds who they think are good.