Was just wondering about this too. Don't have the tech knowledge to weigh in, but the more I thought about it the more I wondered about the magnitude of rapid block propagation (or whatever right term is) issue resulting from small miners having less complete mempools. Sure it's true, but is it a big thing? I know from looking at slow txn attack that a small delay does very little to advantage an attacker (even a 20% attacker only gains a few effective points hash advantage with a 100 second delay, and only on blocks subsequent to the nonstd ones they mine).