What I had in mind:
- Proposers propose headers to the validator.
- The validator commits to all headers he has seen (or at least the highest paying proposal per proposer).
- Upon receiving the commitment, rational proposers disclose their collation body to the validator.
- The validator co-signs and publishes to the VMC the highest paying proposal for which the collation body is available.
The fee F, the subsidy S and the total transaction fees T get paid iff the proposal published to the VMC wins.
if not, then validators still have an incentive to engage in credit scoring
The validator does not get the fee and should still not engage in credit scoring. Imagine a “bluffing” proposer constructs a high credit score over several months with a validator by behaving perfectly. Once the credit score is high enough (say, 99%) the validator stops downloading the full collation body. At that point the proposer challenges and the validator gets slashed. In other words, the credit score needs to continuously be at 100%, i.e. credit scoring is ineffective.