MEV after the tail fork: what MonadBFT removed, and what a trader still pays

MonadBFT’s distinctive claim, as the docs state it, is resistance to tail-forking. In a pipelined leader protocol a leader can sometimes orphan the block the previous leader just proposed. The transactions in that block return to the mempool or vanish from the canonical chain. The people who had already acted on a speculative inclusion are the ones who pay. They bridged, they credited an exchange, they told a customer the payment landed, and then the block was not the block. Monad’s rule is that a leader should not be able to throw that block away. Speculative finality is one slot, and it reverts only if the original leader equivocates. Full finality is the slot after that. If you came because someone said MEV is solved, you were sold the wrong sentence. The tail fork is a consensus failure mode. A sandwich is a market structure. Loose writing calls both MEV. Only one of them was the target of this protocol.

What the paper-shaped claim does for a trader

You may treat a hash from one slot ago as very likely to stay, and you should still be able to say the exception: the original leader equivocated. You may not treat that as a privacy property. You may not treat it as a price property. The block that stays can still contain your swap at a worse price than the quote, with a searcher’s transactions on either side, and the protocol is working as designed when it finalises that block. Finality is about which transactions exist. It is not about whether you enjoyed them. The linear happy-path complexity, the other half of the MonadBFT claim, is about the validator set staying large. A large set does not by itself auction the block to you. It makes it harder for one operator to be the only person who can censor. Censorship resistance and a fair price are different goods. Buy the one the protocol is selling.

The sandwich, in one number

A swap with a public hash and a wide minimum can be seen and traded against before it lands. The searcher buys, your swap pays the worse price, the searcher sells. The profit is the gap between your quote and your floor, minus the searcher’s fees and the risk that they do not get in. A floor of zero hands them the entire output. A floor one percent under a slightly stale mark was reverting honest closes on this desk, which is a different loss: the position stays open, the human taps again, and the second attempt races the first. The floor this desk now starts from is half the spot, so a block moving under you does not trap the coin, and the call is simulated before it is sent. Half of a 50 MON mark is 25 MON. On a deep WETH or USDC pool the pool should not go there. If a searcher can push a 50 MON clip that far, the pool was not deep, and the floor is the most you can lose on that call rather than the amount you expect to lose. Read the new back amount the simulation returns. The floor is permission. The simulation is the estimate. The receipt is the result.

The tail fork was a leader deleting a block. The sandwich is someone trading around your swap. The protocol addressed the first. Your minimum addresses the second, and only up to the number you set.

Which wait matches which regret

One slot of speculative finality is enough for a card you can see and reverse by trading again. It is not enough for a payment you cannot reverse if the rare equivocation happens. The docs are explicit that the rare case requires the original leader to equivocate. Rare is not never. For a close, act on the hash and check the receipt. For a transfer into an exchange, a bridge, or a contract you cannot unwind, wait the two slots. They are 600 milliseconds. The cost of waiting is a pause in the agent loop, not an Ethereum confirmation. The cost of not waiting, in the rare case, is a transaction that was real and then was not, after you have already credited it. Match the wait to the regret. A blanket “wait for finality” on every 50 MON close makes the agent slower than the market for a risk the protocol has already made small. A blanket “the hash is cash” on a bridge makes the rare case into a missing deposit.

The builder still holds the pen

Whoever builds the block orders the transactions that do not conflict with the protocol rules. Parallel execution does not take that pen away. Independent transactions can be ordered in any way that respects conflicts. Dependent ones run in the order the builder picked, with re-execution if the optimistic schedule was wrong. A backrun of a large swap is still a strategy, and on a chain with a large block it is a strategy with room to fit. A user’s defence is a floor that leaves too little to be worth the risk, a size that does not move the pool, and a refusal to send a second copy with a worse floor because the first one felt slow. Private order flow, if a venue offers it, is a different trust: you are trusting the venue not to be the searcher. Read who sees the order. “On Monad” is not a privacy policy. The tail-fork result does not become one just because both subjects appear in essays with MEV in the title.

What to say when the next thread blurs them

If a post says Monad removed MEV, ask which MEV. If they mean a leader orphaning the previous block, the docs support a careful version of that sentence, with the equivocation exception. If they mean your swap cannot be seen, the sentence is false. If they mean the chain is too fast for a searcher, look at the searcher’s location, not at the slot. A searcher in the same process as a builder does not experience 300 milliseconds as a obstacle. You, waiting on a wallet prompt, do. The asymmetry is the product. The protocol did not claim to delete it.