Nice post! It kind of reminds me of 1) Themis, but you ignore the ordering it imposes and 2) the decentralised order flow distributer proposal.
Conditional tipping rules where the transaction tip is only split among the proposers who include a transaction can be used to improve censorship resistance even further. Conditional tipping logic increases the cost of censorship by making colluding equilibria less stable.
It would be interesting to see how this is actually implemented and what the leader’s incentives are to pay the tips.
I also wonder how you would combat the dynamic arising from non-leader validators waiting to the last possible moment to increase their bundle (therefore tip) size, especially in a stake-weighted setup.