Hey @potuz ! Thanks for writing this up. I think the mechanism you describe here is almost the same as Execution Tickets. The difference is that potential builders buy tickets by staking instead of by paying money directly for the tickets. The “buying part” of the two mechanisms is different but the way in which building rights would be allocated is similar, e.g. use RANDAO to allocate building rights randomly with some lookahead.
Execution Tickets has known drawbacks, which you also point out.
- Multi-slot MEV becomes possible because it is known beforehand which builder builds the next block. If builder A knows it builds two slots in a row it could adjust the contents of the first block to extract more MEV in the second one. Multi-slot MEV has negative externalities on users and makes it harder to design applications.
- Block building becomes more centralized since builders need to predict block value.
The fundamental problem with multi-slot MEV in this design is that it is known beforehand who builds the next X blocks. In MEV-Boost builders need to compete again every slot so it is only known who builds the next block just before it needs to be propagated.
The probability of being allocated two consecutive slots in this design depends on the fraction of 0x03 stake any builder controls. I think we cannot rely on that probability simply being very low. The builder market is very concentrated today because the difference in profitability between the top builders and the rest is large. If builders do not collude, they could still find multi-slot MEV opportunities, however, it may then not be rational for ETH holders to stake so multi-slot MEV is not mitigated.