These criticisms are lifted from my replies to this fee escalation tweet.
Payment For Censorship
The escalator pays validators for censoring (txs increase in gas value each block they are censored).
Builders are already incentivized to prioritize private order flow- that’s what we’re trying to fix.
This actually rewards them even more for doing so.
Unstable Competition
Builders are both collectively and individually incentivized to censor txs under fee escalation.
Collectively because they all make more.
Individually because they make their private flow more attractive and secure dominance.
Even users want fast inclusion and simple and predictable mev protection & fee settings, so will also favour an (at first) benign dictator.
So I don’t see how we can consider stable competition to be equilibrium here, from any starting point.
Equilibrium Is Still One Dominant Builder
Even if our starting point is 100 builders all looking only at the mempool, the first one to accept private flow for instant inclusion prioritizes their flow and fills the rest of the block with the most escalated txs.
Any private order flow always means additional profit for the builder because no-one else can bid on it. So they always prioritize it and mempool txs get delayed.
Same situation as before, except now with centralized builders getting paid for those delays.
Gas Price Wars Revisited
As well as this, fee escalation also re-introduces gas price wars, but dragged out over multiple blocks.
If you can see someone else’s tx is going to escalate quicker and beat yours, you cancel yours and resubmit with a faster rate of escalation, and so on…
Elevated Returns From Censorship
The longer a tx is delayed by waiting for the gas to escalate, the more extractable value it will have on average (random walk probability distributions widen over time) which isn’t good if you’re trying to minimize mev losses for the user
Mempool Griefing
I think private order flow prioritization (implicit mempool censorship) is equilibrium.
I also think that mempool griefing of the kind I describe here (explicit mempool censorship) is a very real possibility.
Multiblock Advantage
Pools proposing multiple contiguous blocks can extract more by delaying the most escalating txs until the last block. This gives them a MEV advantage over single block proposers.