Should proof verification happen at the beginning?

[s1na]

In the current token EE prototype we first process txes and produce a set of updated accounts, and only then verify the multiproof and compute the post state root.

This leads to the question of whether this is DoSable. Attackers can craft a block with completely invalid accounts/proofs which imposes a lot of evm bytecode processing, and only after the EE has undergone all the processing can it see that the proof is invalid.

I think this will depend on how the fee market looks like and how the block producer determines whether a block is payable.