Most of the MEV is spam: this paper measures state-invariant Ethereum/L2 txs that burn blockspace and change nothing but fees.
4 comments
If a transaction is state-invariant except for fees, does that mean a rollup could just drop it from history later without breaking anything important, or do people still need the full trace for debugging and fraud proofs?
> does that mean a rollup could just drop it from history later without breaking anything important?
No, not if you still want fraud proofs or practical debugging, since the tx data and execution trace are what let you reexecute and explain a bad state root, even when the tx is fees-only at the end state. What exactly survives pruning here, full calldata, traces, or just a receipt summary, and do these state-invariant txs still have to stay around for the DA window on Optimism/Base?
> does that mean a rollup could just drop it from history later without breaking anything important?
Not during the fraud window, and not if you care about reconstructing what happened later. For Optimism/Base the compressed batch data has to hang around long enough to let anyone reexecute and challenge, and a receipts only summary is not enough for that. After the window, sure, an archive node can prune more aggressively, but that is a separate storage policy, not something state invariance buys you by itself.
The paper’s point is more annoying than “safe to delete”, it is that these txs still consume DA and execution even when they end up changing nothing.
The “safe to delete” angle is the red herring here. Fees-only doesnt mean side-effect-free, it still burns proposer time, batch space, and gives you an ordering signal that can matter before the final state settles.
And no, archive pruning after the fraud window isnt some magic cleanup, it just shifts the mess onto whoever wants to debug or audit later. The ugly bit is these txs can be economically useful spam even when they are computationally boring, which is a much less tidy problem than “just drop them”