But the point was that is not a stable equilibrium because the profits are not accruing proportional to hashrate due to unequal propagation or orphan rates (i.e. wasting more hashrate). Thus over time the hashrate concentrates until there is a 33 - 50% attack and then eventually the cartel can set any block size it wants with a 50% attack.
While it's true that bigger hashrates and better network connectivity might result in centralization (eventually leading to a 33% or 50% concentration), the problem of proportionality isn't specific to the question of unlimited block size or the fee market in general. It also occurs for block rewards (
https://blog.ethereum.org/2014/07/11/toward-a-12-second-block-time).
In the short run, I don't see how this problem would affect the equilibrium. It's an expected behaviour of markets that the suppliers don't have the same conditions and cost structure, which is why their individual supply curves all look different. Together they build the market supply, so that an equilibrium can finally be reached.
What makes the present case (unlimited block size vs. tx fees) really different from other markets is not the centralization aspect (economies-of-scale exists in a lot of markets), but the fact that the supply curves are interdependent.
It's a bit like a group of professional fishers who are all fishing in the same lake and selling their fish to the same consumers. The amount of fish produced by any single fisher can contribute to overfishing and diminish the supply of the other fishers as well.