I believe Danny's point was that it invalidated the midstate to change it.
Oh, wait, I'm not going to prove someone wrong. This thread become quite nice mix of math and conspirology, that's very fine, let's continue

If there exist some algebraic attack that breaks two rounds of SHA256 (and OP is hunting for it with statistics), it's quite possible that the variant of such attack will bypass midstate and break 3 rounds, from 640 bit header to nonce. So the next suspicious thing are unusual block versions in the blockchain.