Thanks for this! There are some similarities with this proposal. It’d be interesting to get your thoughts on how the two differ!
It’s a shame – in the case of multiple VRF ‘winners’ racing to be the first to submit to L1 – that there’s no way for ethereum to refund (or not charge) the losing submitters for their wasted calldata (and maybe eip4844 blob) data submission costs. It’s not like the vast amounts of calldata (of a losing rollup) are ever needed by the eth network in future, and the data won’t even have been used in the body of the function, if it’s designed to fail on ‘line 1’.