Um, maybe I misunderstood what you said or the proofs, but isn't there a much simpler way to reject the proof of superiorness of the WMA, namely that in the first proof is, as you say, the worst case scenario.... but the second proof is _NOT_. It uses expected value rather than worst case, right? So it's not so much "worst case analysis does really weird things" but "not only that, the two proofs aren't even validly comparable, because one is worst case analysis and one involves expected value analysis"

Or did I horribly misunderstand either the second proof or your argument against it? (if so, I blame lack of sleep)