Thanks CryptoWhite,
Yes, both models have two separated roles, but the scope and the implementation of PBS are very different.
PBS is focusing on mitigating MEV centralization risks on L1 validators while in PoE the objective is to provide a permissionless model to enable decentralization on zk-rollups while solving issues like managing the gap between network confirmation and ZKP finality, which today is only achieved through a centralized operator.
Yes, the Aggregator mechanism for sure can be improved. As you describe there is a waste of computation since only the most efficient will get the reward, but this probably will lead to a reduced number of competing Aggregators and it would not be a lot of total computation.
We are also considering the option to introduce some randomness in the ZKP acceptance or even to distribute the reward between close proofs in time. Ideas are very welcome here…
I think I don’t understand what you mean here, the Revert will be the result for late ZKP for the same set of batches (virtual states). The idea is that finality proofs validate a set of batches and not a single one, so we can optimize the L1 gas for the costly ZK proofs.