Can we consider assigning validators as includers proportional to their stake, as is already done for proposers? The reason is that income from inclusion duties may discourage validator consolidation. To the extent that keeping stake split across smaller validators offers more of this income, it seems rational for an operator to do so. This could undermine consolidation, which is useful for fast finality.
To see why the inclusion role may carry income at all, first consider what a spot in an IL is worth to a transaction. The marginal value attributed to having a transaction in an IL can be loosely defined as the reduction in priority fees that still allows inclusion. In a non-full block, the sender of an IL transaction will not need to pay any priority fee. As long as the block producer is better off accepting an IL transaction than it would be by filling the block with dummy transactions in order to censor it, the transaction will get included. The marginal value to a transaction of the IL is thus affected by any frictions that transaction inclusion may incur, for example by making the block slower to propagate or introducing legal questions. Of course, the biggest frictions are incurred by transactions that would otherwise be censored for more indirect reasons, for example to force a liquidation and extract MEV, or even more “external” reasons, such as sanctions. But my point here is to highlight the sliding scale of “censorship” that may not involve outright refusal of inclusion, but rather hinge on the size of the priority fee.
Furthermore, the transaction may for that matter simply not reach the self-building solo staker in time through ordinary transaction gossip, such that the IL path is the only viable option for inclusion in that slot, thus giving that path value. And of course, a centralized builder set would do best to use any pricing power it has to not include transactions outside of the IL that pay too little in priority fees, even when there is room for them.
At the same time, we must suspect that priority fees will compress already with FOCIL. The compression is nevertheless likely to remain partial, if the ILs, each capped at 8 KiB, cannot hold every transaction that would benefit from them, especially once overlaps between them are taken into account. Thus, while the marginal value of a spot in the IL may be highest for transactions that would otherwise be censored for external reasons, more transactions may ultimately derive value from being in an IL.
It stands to reason that some of this value may accrue to includers, in the form of out-of-band payments, which includers base their inclusion decisions on. Competition from includers willing to list transactions for free may limit such payments, but senders excluded from, or uncertain of, those free options could still pay for reliable inclusion. The price of a spot may then reflect the scarcity of IL space and the value senders place on certainty. Competition among includers need not drive this price to zero if they cannot jointly accommodate every transaction that derives some value from a spot. Some of the value that would otherwise accrue to builders, and through their bids to proposers, may thus accrue to includers instead, shifting income from a role assigned in proportion to stake to one assigned per validator. To extract the most value, includers may need to expand their services to also negotiate the final payment to the block producer on behalf of senders in the case that the block is (potentially close to being) full, a payment which may also be made out-of-band.
Now, we must assess the importance of consolidating the validator set against the potential benefit of giving smaller validators a higher chance per unit of stake of becoming includers, thus potentially increasing censorship resistance. Uniform selection favours smaller validator balances, however, not necessarily smaller operators: a large staking service provider (SSP) can retain the same advantage by keeping its stake split across validators. Indeed, the more others consolidate, the more includer seats fall to those who do not. If 70% of the stake were consolidated into validators of 2048 ETH, the remaining 30% held in validators of 32 ETH would occupy about 96.5% of all includer seats in expectation. Were that 30% controlled by a party intent on censorship, it would hold all 16 seats in more than half of all slots, something that would practically never happen under stake-weighted selection. Uniform selection may thus weaken the very censorship resistance it appears to strengthen, for such a party has every reason to keep its stake split. This is not to deny that solo stakers holding 32 ETH would also gain seats, and thereby strengthen censorship resistance. The difficulty is that the protocol cannot tell their validators apart from those of a party that chooses not to consolidate in order to gain pricing and censorship power.
Under the assumptions outlined above, the inclusion role can thus give an SSP a reason not to consolidate, to be weighed against the modest operational savings that consolidation offers. Since the main benefits of consolidation accrue to the network as a whole rather than to the operator, even a modest income from inclusion duties could tip this balance. The income for includers may indeed be very small, because priority fees for openly propagated transactions may very well compress without accruing to includers. Some SSPs are already consolidating: perhaps due to a belief that it will be difficult to extract much value from the inclusion role, for operational simplicity, due to social pressure, or a wish that Ethereum prospers.
The social pressure to consolidate may be enough. What I am asking for is a closer discussion of our assumptions. The social pressure to make unprofitable choices is never as strong as the social pressure that can be achieved when the choice is neutral. A competing SSP may keep its validator set unconsolidated to attain a high probability of having at least one includer in any given slot, offering senders a reliability that no includer drawn at random can match, and taking payments out-of-band. Or such services may emerge in a more general fashion (analogous to MEV-Boost). This would make an SSP with a consolidated validator set less competitive.
We must finally also consider that the includer role is rather versatile and may come to be expanded in the future, giving even better opportunities for includers to extract value. If we pursue an encrypted mempool, the ILs can come to facilitate inclusion also for transactions that currently rely on private submission. If we pursue an expansive multidimensional fee market, UPIL may be adopted to preserve censorship resistance, allowing the ILs (together with the ranking fee) to force inclusion also in full blocks. We may also in the future allow includers to get paid directly by senders in-protocol, solidifying IL income also for validators that do not wish to rely on out-of-band payments. Admittedly, the balancing act here must also account for the actual value we assign to a consolidated validator set, in terms of reducing consensus overhead and making faster finality easier to achieve, and thus needs broader input from the ecosystem.
