Why Cross-Domain MEV Makes Fairness Intractable

A searcher spots a large swap on Uniswap on Ethereum. They front-run it with a flash loan, then execute the same swap on a faster L2 like Arbitrum via a cross-chain bridge like Across, capturing the price impact before the original Ethereum transaction finalizes.\n- Key Insight: Latency arbitrage between L1 and L2.\n- Result: User gets a worse price; profit is extracted across domains.

LayerZero's generic messaging enables complex cross-chain intents. A relayer can observe a pending transaction on Chain A, compute a profitable dependent action on Chain B, and guarantee its execution first by controlling the message delivery order.\n- Key Insight: Relayer as a centralized sequencer across chains.\n- Result: The "oracle" and "relayer" roles become a single MEV extraction point.

UniswapX uses fillers to execute intents off-chain. A malicious filler can see an intent to swap A for B on Ethereum, but instead of filling it, they front-run the destination swap on a DEX on Polygon, profiting from the impending price move.\n- Key Insight: Intent-based systems leak information to solvers.\n- Result: Fairness depends entirely on the filler's honesty, a known broken model.

Bridges like Multichain (formerly Anyswap) rely on oracle networks for cross-chain consensus. A validator can delay or reorder price feed updates between chains, creating arbitrage opportunities in bridged asset pools (e.g., USDC.e on Avalanche).\n- Key Insight: Cross-chain state consensus is a new MEV vector.\n- Result: Bridge security directly determines cross-domain fairness.

During an Optimistic Rollup's challenge period (e.g., Arbitrum's ~7 days), a sequencer can theoretically rewrite history for a previously submitted batch if they discover a more profitable transaction ordering, forcing a reorg.\n- Key Insight: Delayed finality enables retrospective MEV.\n- Result: Fairness isn't just about the present; it's about the past week.

Projects like Astria and Espresso aim to provide shared sequencing across rollups. However, if each rollup can choose its own sequencer set, a searcher can corrupt one sequencer to gain an advantage in cross-rollup arbitrage, fragmenting the security model.\n- Key Insight: Shared sequencing without shared security.\n- Result: The weakest sequencer determines cross-domain fairness.

Transactions are no longer atomic across domains. A user's swap on Ethereum and bridge to Arbitrum are separate events, creating a multi-domain sandwich attack surface.\n- Latency arbitrage: Extractors exploit the ~12s delay between L1 finality and L2 state updates.\n- Fragmented execution: MEV is no longer a single-block puzzle; it's a multi-chain, multi-venue game.

MEV extraction is balkanized. LayerZero relayer, Across relayers, and L2 sequencers are now independent, competing extractors.\n- No shared mempool: Cross-domain intents are opaque, preventing a unified fair ordering solution like MEV-Boost.\n- Cartel formation: Dominant relayers (e.g., for major bridges) can become centralized profit hubs, extracting rents across the entire flow.

The only viable path forward is to abandon transaction-based models. Protocols must design for declarative intents.\n- UniswapX & CowSwap: Pioneer the model—users submit desired outcomes, solvers compete cross-domain.\n- Protocol Role Shift: Become intent originators and verifiers, not transaction sequencers. Cede execution complexity to specialized networks.

Fairness requires a new primitive: a cross-domain block space market. This isn't just about ordering, but atomic scheduling across non-atomic domains.\n- Shared Sequencing (e.g., Espresso, Astria): Attempts to create a unified sequencing layer for rollups, but faces L1 finality constraints.\n- Economic Alignment: Future systems must use cryptographic proofs (like ZK) to make cross-domain violation claims punishable, aligning sequencer incentives.

With intents, the value shifts from transaction ordering to information asymmetry. The entity with the best cross-chain state view wins.\n- Oracle MEV: Extractors front-run based on privileged price feed updates across chains.\n- Privacy Leaks: Intent submission patterns reveal user strategies, creating new extractable signals before execution.

1. Design for Intents: Expose solver networks via standard APIs (like UniswapX).\n2. Assume Extractor Sovereignty: Build verification, not control. Use ZK proofs for cross-domain settlement.\n3. Own the Data Layer: Integrate with fast, finality-aware oracles (e.g., Chainlink CCIP) to mitigate information-based MEV.