Edit: I’ll keep this post up for anyone else that wants to read. But the questions I had below were largely due to two misunderstandings:
- In the proposed design, the proposer is allowed to keep the MEV difference between the bid floor (what I called the attested bid) and the bid he selects for the block (what I call the proposed bid).
- While the builders are not directly incentivized to gossip competitive bids (what I called participating in the attester auction), they’re not dis-incentivized from doing this either. The value of bids from builders are already public today, so builders are not giving up information by gossiping this.
Original post below:
Forgive me if I’m misunderstanding, but I feel like something needs to be clarified here.
- non-colluding builders: If one of the builders capable of extracting a given piece of MEV defects by bidding there is no benefit for any of the builders to bid late. If anything, late-bidding builders risk not having bids reach the proposer on time.
It’s worth being explicit about the fact that the builder who submits the bid with the highest base fee (as chosen by the attesters D seconds before the slot), does not necessarily win the right to build the block. In other words, another builder could still bid higher after the cutoff & win the auction. This must be the case, as awarding the proposal rights to the winner of the attester auction removes the whole purpose of having a proposer. It would also result in increased re-org probabilities. The whole purpose of having a gap D is to stop split views of the top bid from creating forks.
But having a delay means there are really two auctions:
the attester auction: where the winning bid (the attested bid) is selected by attesters
the proposer auction: where the winning bid (the proposed bid) is selected by the proposer
We would expect these to often be different bids as the proposer auction takes place several milliseconds after the attester auction which usually corresponds to more MEV. There is incentive for proposers to engage in off-chain agreements with builders here as proposers can extract the value difference between the proposed and attested bids.
Suppose we reflect this on chain by allowing the proposer to submit both bids. In a similar way, the attesters enforce that the attested bid matches at least what they see as the attested bid & the attested bid is burned. The proposer will want to wait a small delay past the cutoff before selecting the attested bid to ensure that gossip latency won’t cause them to under shoot the attested bid. But we also allow the proposer to submit the proposer bid & extract this difference. This could remove much of the incentive for the proposer and builder to collude. By including the attested bid, the chain can know who won the attester auction. It’s also worth noting that these bids could be the same.
This won’t work because the proposer could create their own bid and say they won the attester auction 
With that out of the way, let’s revisit this statement:
If anything, late-bidding builders risk not having bids reach the proposer on time.
Here we are incentivizing the builder to participate competitively in the attester auction by assuming that they risk losing the proposer auction if they don’t submit a competitive bid by the cutoff. But that risk is a function of the gap between D and the gossip bid delay. The larger this gap, the lower the risk, and this gap can be artificially increased by using out of protocol relays that enable low-latency connections between the proposer & builder. It’s also just generally sub-optimal to rely on optimal tuning of timing parameters for security.
It would be better if there were economic incentives for winning the attester auction. One way to do this is to award the winner of the attester auction in this slot some fraction of the MEV burned in the next slot.