Introduction
MEV is a vast and complex field; in this article I want to gather some of the key points from the last 6 months and provide a snapshot of where we are and where we are going.
The latter half of 2025 marked a decisive shift toward "architectural ossification" embedding MEV management directly into L1 and L2 protocols.
This period saw the formalisation of Enshrined Proposer-Builder Separation (ePBS) for Ethereum's "Glamsterdam" upgrade, the operationalisation of Trusted Execution Environments (TEEs) for L2 sequencing, and the maturation of "Based Sequencing" as an alternative to centralised sequencers.
The catalyst was recognising MEV as the "dominant limit to scaling blockchains."
High-throughput chains encountered an "MEV spam wall" where arbitrage bots consumed over 50% of block space while contributing negligible fees. This forced a design change:
- one path tames MEV through cryptographic privacy and hardware (TEEs), while
- the other democratises it through L1 integration (Based Rollups).
The Consensus Layer: ePBS and the Glamsterdam Upgrade
EIP-7732 (ePBS) removes reliance on trusted mev-boost relays by allowing builders to interact directly with the consensus protocol.
The slot is split into a bidding phase and payload revelation phase, with the protocol enforcing delivery through slashing.
The Free Option Problem
A landmark paper revealed that ePBS gives winning builders a "free option" to withhold payloads if market conditions shift unfavourably during the roughly 8 second revelation window.
Builders engaged in CEX-DEX arbitrage may choose to forfeit the block reward rather than execute a "toxic" trade.
Key findings:
- Under normal conditions we see ~0.82% missed block rate
- On high volatility days we have up to 6% missed blocks (over 80% of all misses)
- Paradoxically, as on-chain markets improve, builders rely more on CEX-DEX strategies, making consensus even more fragile
Mitigations under debate:
- Window shortening (2s reduces exercise probability by 77%)
- Increased penalties (raises barriers for smaller builders)
Core developers favour a hybrid approach, though shorter windows inadvertently strengthen geographic centralisation.
The "Based" Revolution: L1 Sequencing and Preconfirmations
Based Sequencing requires that L2 rollups should rely on Ethereum L1 proposers for transaction ordering, inheriting L1's liveness and censorship resistance. Here "preconfirmations" solve the latency trade-off.
Taiko's Preconfirmation System
Taiko's Phase 1 (August 2025) utilised whitelisted preconfirmers providing cryptographic inclusion guarantees, reducing confirmation times from 12 seconds to about 2 seconds.
Phase 2 will make the preconfirmer set fully permissionless, with community consensus suggesting 20% of L1 stake should opt-in for a reliable service.
Nethermind's Surge Framework
Surge achieves "Gigagas" performance (1 billion gas/second) through parallel execution optimisations.
It launched as a Stage 2 rollup with permissionless fraud proving and no security council override.
ENS validated the model by migrating its Namechain L2 to Surge, citing "Ethereum-native settlement" and censorship resistance as non-negotiable requirements.
Hardware-Enforced MEV: TEEs and Unichain
Unichain became the first L2 to mandate block building within a TEE, fundamentally altering MEV dynamics.
Key features:
Encrypted Mempool: Transactions are decrypted only within the TEE, preventing sandwich attacks
Priority Ordering: Deterministic ordering by priority fee eliminates spam incentives
Flashblocks: Partial blocks streamed every 200-250ms for near-instant confirmation
Revert Protection: Failed transactions are excluded rather than charged
Centralisation risk: Reliance on Intel TDX creates a single point of failure. Compromised attestation keys or side-channel attacks could invalidate chain guarantees.
Empirical Market Dynamics
The MEV Spam Wall
Flashbots research found that when Base increased throughput by 11 million gas/second, MEV spam bots consumed nearly all new capacity, over 50% of gas usage but less than 10% of fees. This justifies Unichain's design choices.
The Geography of Block Building
Network latency determines PBS auction success. Builders must colocate near CEX data ingress points and relay infrastructure. The "Ashburn-Tokyo" axis dominates winning blocks a physical centralisation that resists protocol-level decentralisation efforts.
Latency in Common-Value Auctions
Research using Black-Scholes models shows the value of being the last bidder increases non-linearly with volatility. Preventing monopolist builders may require multi-block auctions or frequent batch auctions.
Shared Sequencing and Interoperability
Espresso Systems advanced toward Mainnet 2, featuring HotShot consensus for high-throughput, fast-finality sequencing across multiple rollups.
Their "Shared Sequencing Marketplace" allows rollups to auction sequencing rights, enabling atomic cross-chain transactions.
Regulatory Intersection
The ESMA TRV Risk Analysis (July 2025) provided the first regulatory taxonomy of MEV:
- Benign MEV: Arbitrage and liquidations (market-neutral)
- Toxic MEV: Front-running and sandwich attacks (market abuse under MiCA)
This creates regulatory incentives for L2s to adopt MEV-preventing architectures.
In conclusion
The May–November 2025 period represents "Architectural Ossification" in MEV research.
The Consensus Layer hardens with ePBS while grappling with the Free Option Problem.
The Execution Layer has split:
- Hardware Path (Unichain): Prioritising UX through TEE trust
- Native Path (Based Rollups): Prioritising decentralisation via L1 integration
The recognition that scaling limits are now economic efficiency and the speed of light, not just bandwidth, will define the 2026 research agenda.
Originally published on: academy.extropy.io
Top comments (0)