Yes, good points. I was thinking of Bitcoin as an asset and not the numéraire, which to my mind changes the situation a bit - the problem suggested is that some 'undesirable' people will end up with too many Bitcoins and use them to manipulate the market. The Coase theorem points to this being as unlikely as it is for any other asset class. The bad actors will eventually get bought out if the putative damage they're doing is great enough - and, compared to the size of transaction implied, the costs will be small.
I think a part of what you're missing is that (says my understanding, which could certainly be flawed...) transaction costs in Coase's model include the costs of enforcing whatever contract is used to prevent whatever badness.
