Ethereum News
Optimism sequencer centralization: myths vs facts for rollup security

The single operator role in transaction processing
Every production Ethereum Layer 2 in 2026 runs a single sequencer operated by one entity. This includes Arbitrum, Base, OP Mainnet, zkSync Era, Linea, and Scroll. A sequencer receives transactions, chooses an order, executes them against the rollup state, and batches the results for submission to L1. The Optimism Foundation operates the centralized sequencer for OP Mainnet to prioritize performance and simplicity. This model allows the sequencer to process transactions every two seconds and achieve near-instantaneous execution. The sequencer also collects, validates, and sequences transactions. This centralization allows for lower user costs by aggregating and compressing transactions.
The sequencer performs four jobs in sequence on every block. It accepts incoming transactions, chooses an order, executes them against the current state to produce new state, and batches the result for submission to L1. The first three jobs happen in milliseconds. The L1 submission step happens on a slower cadence measured in seconds or minutes. In March 2024, the Ecotone upgrade introduced EIP-4844 blobs to Optimism. This change reduced data storage costs and helped lower transaction fees by over 90%. As of November 2025, Arbitrum One holds $16.63 billion in bridged assets. Base Chain holds $10 billion, while OP Mainnet holds $6 billion.
Liveness and censorship risks in production
A centralized sequencer concentrates downtime, censorship, and MEV extraction into one piece of software. On June 25, 2026, Base halted block production for 1 hour and 18 minutes because a consensus failure sequenced an invalid block. This outage froze deposits, withdrawals, and swaps. User funds remain safe because the L1 state does not change, yet no new transactions can receive "soft confirmation" during such periods. Coinbase also filters OFAC SDN addresses through the Base sequencer. In February 2025, Base experienced a 33-minute outage because of a failed sequencer handoff during high activity. Arbitrum also saw an outage in December 2023 during the Inscriptions traffic surge. While the L1 force-inclusion mechanism protects against permanent censorship, the delay of 12 to 24 hours on most major rollups provides no protection for a trader facing a margin call due in 40 minutes. You should monitor liveness anomalies if you use these chains for time-sensitive transactions.
MEV exposure and private mempools
The sequencer mempool on an OP Stack chain is private and visible only to the sequencer. This design limits the visibility that front-running strategies on public mempools depend on, but the sequencer still sees every pending transaction. In a public mempool, anyone can see pending transactions, which creates MEV opportunities. On an OP Stack chain, the sequencer remains the only party with visibility. Regarding profit, Optimism shows a median profit of 0 and a mean profit of 0.0014 per block. Arbitrum shows a median profit of 0 and a mean profit of 0.319 per block. Polygon shows a median profit of 0 and a mean profit of 53.70 per block.
Since the sequencer controls ordering, it can insert, reorder, or sandwich user transactions for profit. In April 2025, Arbitrum launched Timeboost to auction an express lane. An empirical study found that three entities won 99.74% of all auctions. Approximately 30% of these express-lane transactions reverted during the study period.
| Configuration | Centralized Sequencer | Based Rollup |
|---|---|---|
| Operator Count | One | Ethereum Validators |
| Latency | Sub-second | 12-second slot time |
| MEV Destination | Sequencer Operator | L1 Validators |
Decentralization paths and architectural trade-offs
Decentralization paths include shared sequencing and based rollups. Based rollups, such as Taiko Alethia, delegate sequencing to Ethereum L1 validators. This approach maximizes alignment with Ethereum but inherits the 12-second Ethereum slot time for hard ordering. Shared sequencing, like the Espresso Network, uses a set of operators to sequence transactions for multiple rollups. Espresso launched its Mainnet 1.0 in late 2025 and provided confirmations for integrated chains like ApeChain and Celo by early 2026.
The industry continues to weigh these architectural choices against the need for speed. Developers choose between the speed of a single operator and the security of a distributed set. Can the market achieve full decentralization without sacrificing the sub-second latency users expect?