Ethereum News
Centralized sequencers and the shift to proof-based security
Every major Ethereum Layer 2 rollup still runs a centralized sequencer operated by a single entity as of October 2026. This includes Arbitrum, Base, OP Mainnet, zkSync Era, Linea, and Scroll. These single-operator setups create risks regarding censorship, MEV extraction, liveness, and transaction delays. The Linea team paused its sequencer in June 2024 after a Velocore DEX exploit caused $6.8 million in losses, an incident that sparked debate about whether a single entity should hold the power to halt a decentralized network. While the proof systems prevent the sequencer from stealing funds, the operator can still decline transactions or reorder them to capture arbitrage.
A sequencer performs four primary tasks. It receives transactions from the L2 mempool and decides their order. It assembles these ordered transactions into blocks and executes them against the L2 state. It then compresses these blocks into batches and posts them to Ethereum L1 as EIP-4844 blob data every 60 to 120 seconds. Finally, the sequencer submits a Merkle root of the new L2 state to the rollup’s L1 contracts. Arbitrum Nitro can process over 40,000 transactions per second internally before it posts compressed batches to Ethereum. Between August and September 2026, Base generated approximately $5.3 million in sequencer revenue, while Arbitrum generated $0.47 million.
The risks of this centralized model remain concrete. A single operator can exclude transactions, which creates censorship risks for users and protocols. The operator also sees every transaction before it reaches finality, which allows for MEV extraction through sandwich attacks or front-running. If the sequencer goes offline, the entire chain stops and DeFi protocols freeze. This dependency on a single point of failure was evident when Coinbase experienced a downtime event that halted the Base chain in February 2025.
Optimism and the Fault Proof Transition
Optimism is moving toward Stage 2 status through the deployment of fault proofs. These proofs allow any actor to participate in the dispute game to challenge invalid output roots. The system uses bonds to cover the anticipated cost of posting a counter claim and to deter spamming invalid claims. A single claim can face multiple counter-claims from different actors within one dispute game, which can have a maximum depth of 73 steps. The Guardian role can override the system by blacklisting games or reverting to a permissioned system. This reduces the trust assumption because users only need to believe the Guardian will not intervene.
Normal users do not typically run the op-proposer to propose output roots, as they can instead use direct calls to the DisputeGameFactory or the create-game subcommand of op-challenger. The roles of proposer and challenger remain open to everyone. Optimism also plans to integrate shared sequencing infrastructure, such as Espresso, in 2026 to achieve production rollout. Espresso currently runs a permissioned set of approximately 100 geographically distributed nodes on HotShot consensus. This network has processed over 20 million transactions and secures more than $300 million in total value. This development comes after Astria, which built a similar system on Celestia, shut down in December 2025.
| Feature | Optimistic Rollups | ZK Rollups |
|---|---|---|
| Verification Timing | After 7-day challenge window | Immediate via cryptographic proof |
| Trust Model | Honest challengers must act | Cryptographic certainty |
| Computational Cost | Low (only if disputed) | High (per batch) |
| Withdrawal Speed | 7 days | Proof generation time |
zkSync and the ZK Stack Architecture
The zkSync ZK Stack introduces a StateTransitionManager to manage the deployment and upgrades of ZK Chains. This upgrade allows Matter Labs to deploy ZK rollup instances that function as independent chains. In early 2026, ZKsync Era and other Elastic Network chains temporarily migrated to settling directly on Ethereum L1 to facilitate this protocol upgrade. This migration allows the deployment of the Atlas upgrade and new components like ZKsync OS and Airbender. The StateTransitionManager allows the owner to define how ZK Chains initialize and upgrade.
The StateTransitionManager manages the deployment of ZK Chains and can freeze or unfreeze specific chains. The Admin facet manages feature settings like the token multiplier and the Pubdata pricing mode. The Executor handles the state-finalization functions. The Admin can also set the Pubdata pricing mode to either Rollup or Validium mode. In Validium mode, no pubdata is committed to the Executor. The owner can define how ZK Chains are initialized through the StateTransitionManager.
The ZK Stack provides different security properties based on the selected mode. Users who require immediate settlement and strong cryptographic guarantees often prefer ZK-based systems. If you are a developer building for compliance-sensitive flows, you should document a force-inclusion path.
Arbitrum Security and the BoLD Upgrade
Arbitrum is rolling out BoLD permissionless fraud proofs throughout 2026. This development advances the chain toward Stage 1 and 2 maturity. To mitigate sequencer-driven censorship, Arbitrum developed a Censorship Timeout feature. This feature lowers the force-inclusion threshold when unexpected delays or outages occur. On October 2, 2026, the Arbitrum Security Council executed an emergency action to pause settlement on Ethereum due to a bug in the BoLD one-step proof. This action allows the council time to deploy a fix while the chain continues to process transactions normally.
The Security Council used the ability to raise the Stylus activation gas requirement to make new contract activations prohibitively expensive. This action also prevented the deployment of hand-crafted WASM programs that could degrade performance. The one-step proof guard allows anyone to show the guard two conflicting answers to a single step of an open challenge. If the one-step proof accepts both, the guard can put Arbitrum One’s settlement to Ethereum on hold.
The sequencer on Arbitrum handles transaction ordering through Priority Gas Auctions. The sequencer evaluates the order in short rounds that run several times per block. If a transaction remains in the queue, it gains a priority boost to ensure it eventually gets included. This boost moves a transaction’s position in the queue but never changes the fee it pays. The sequencer uses the arrival timestamp to break ties in the priority queue. Can the industry move toward full decentralization without sacrificing the latency required for modern applications?
