Test fairness and atomicity claims where they are most likely to fail

Established background

The following is general field knowledge, stated cautiously.

Maximal extractable value (MEV) is value that whoever controls transaction ordering and inclusion can capture beyond ordinary fees — for example by trading ahead of or around a visible pending transaction. Mitigations proposed in the field fall broadly into hiding transaction contents until ordering is fixed (commit-reveal, encryption), constraining the ordering rule itself, and auctioning or redistributing the value.

Composability is the ability of one transaction to use several contracts together and either succeed entirely or fail entirely. On a single chain this is the default; across separate rollups it is not, because each orders and settles independently. Approaches to restoring it typically introduce shared sequencing or cross-domain coordination, which moves trust onto new components.

In both cases the guarantee is conditional on who controls ordering and timing — which is why the assumptions come first.

My results

• FairFlow Protocol (preprint, 2023) proposes an MEV-mitigation mechanism for Ethereum that combines commit-reveal ordering with economic incentives, aimed at more equitable treatment of transactions.
• Towards Universal Atomic Composability (preprint, 2023) proposes a formal model of atomic composability for multi-rollup environments on Ethereum and analyses the conditions for cross-rollup atomic execution.
• Centralized Intermediation in a Decentralized Web3 Economy (preprint, 2023) examines how intermediaries accrue and extract value — including through information asymmetry and ordering — despite decentralised infrastructure.

All three are proposals and analyses; none has been peer reviewed, and none is offered here as evidence that a deployed system is fair or atomic.

Open questions

• Which ordering assumptions does a commit-reveal design still depend on — for example, whether the party that sees reveals can also delay or censor them?
• Under what sequencer failure and delay models does cross-rollup atomicity degrade, and how quickly?
• When a mechanism reduces one form of extraction, where does the value move instead?

Illustrative example — not a client engagement. A cross-rollup swap design claims atomic settlement. The study models two rollups with independent sequencers, bounded message delay and one sequencer that may stall. The search finds an execution in which the first leg commits and the second times out after the refund window closes. The report gives the replayable trace, the timing assumption it violates, and the condition under which atomicity holds.