I have comments on the benefits - but will leave those for another day to focus on the “common misconceptions”.
The two issues are actually the same, and they’re being brushed under the rug when they’re a much bigger issue than they’re made out to be.
Collusion doesn’t have a cost and doesn’t require not bidding like the example shows. Collusion only requires not bidding before the payload observation deadline. If a builder defects to set the bid floor, the other builders can bid at that bid floor after the payload observation deadline. Bids are still accepted after the payload base fee snapshot so defecting doesn’t give you any benefits such as guaranteeing you’d win the block.
This results in collusion being a stable dominant strategy for builders – builders get no benefit by bidding early, but they lose ETH in the case where the validator is offering mev-burn refunds. You really only have 2-3 builders with the ability to build “full value” cex-dex blocks right now - and bidding late is minimal effort, any builder could implement it in an hour if that. It seems obvious to me that both builders will immediately start doing so, if not on their own, the second a validator makes the offer to refund some portion of the mev-burn.
Finally, the historical estimation is no good. You can look at current graphs and say that the block value is bid 80% of the total bid by the deadline. But that’s because there’s a very weak incentive to bid late now (hiding your bids from competitors). This will only get worse as a real incentive to bid late appears (maximizing your validator refund). At best this should be looked at as a generous upper bound.