Skip to content

Guide · Rollups

Sequencer centralisation: who orders your L2 transactions

Why most rollups still run a single centralized sequencer, what censorship and downtime risk that creates, and the shared-sequencer and based-rollup approaches aiming to change it. Not financial advice.

By Etherelysium Editorial TeamUpdated 18 November 20267 min read

Sequencer centralisation: who orders your L2 transactions
Share X LinkedIn

Most rollups in production today rely on a single sequencer — one party, typically the team operating the rollup — to receive user transactions, decide their order, execute them, and batch the results for posting to L1. That single point of ordering is a practical engineering choice that got rollups to market fast, and it is also the clearest centralization bottleneck in most L2s today, independent of whether the rollup is optimistic or ZK.

What a sequencer actually does

A sequencer performs three related jobs: it accepts incoming transactions from users, it orders them into a sequence, and it executes that sequence to produce the rollup’s next state. On most rollups today, the same party does all three, and users have no say in the ordering beyond whatever rules the sequencer’s software applies. The sequencer then compresses the batch and posts the data to L1 — increasingly into a blob, as described in the guide on blob fees — along with, depending on the rollup’s design, a state root or a validity proof.

Why a single sequencer is a centralization risk

Censorship

A sequencer that controls ordering can also choose to delay or exclude a specific transaction indefinitely. Most rollups mitigate this with a “forced inclusion” mechanism that lets a user submit a transaction directly to an L1 contract if the sequencer refuses to include it, but that path is typically slower and more expensive than the normal route, and its practical availability and cost vary by rollup.

MEV extraction

Whoever controls transaction ordering controls the most direct lever for extracting MEV — the companion guide on MEV in plain English covers the mechanics in more depth. A centralized sequencer can, in principle, order transactions to its own advantage, front-run its own users, or sell ordering rights to third parties, with far less of the competitive, multi-party structure that L1’s proposer-builder separation introduces.

Downtime as a single point of failure

If the sequencer goes offline, the rollup can stop processing new transactions entirely, even though the underlying L1 and the rollup’s already-settled state are completely unaffected. This has happened on various rollups during outages; the rollup’s data already posted to L1 remains safe, but new activity simply halts until the sequencer — or a fallback mechanism — comes back.

Paths toward decentralization

Several approaches aim to reduce reliance on a single sequencer, each with different maturity levels and tradeoffs:

  • Shared sequencers: a separate network of nodes that sequences transactions for multiple rollups at once, aiming to standardize fair ordering and enable cross-rollup transaction composability rather than leaving each rollup to sequence in isolation.
  • Based rollups: instead of running a dedicated sequencer at all, transaction ordering is delegated to Ethereum’s own L1 block proposers, inheriting L1’s existing decentralization for the ordering step at the cost of tying the rollup’s sequencing cadence to L1’s.
  • Sequencer rotation or auctions: rotating the sequencer role among a known set of operators, or auctioning the right to sequence for a period, which spreads out control without necessarily achieving full permissionless participation.
  • Forced-inclusion and escape-hatch contracts: an L1-level backstop that lets users bypass an unresponsive or censoring sequencer, which does not decentralize ordering itself but limits how much damage a single sequencer can do.

None of these is universally adopted yet, and each rollup’s actual implementation — rather than its stated roadmap — is what determines how much protection a user actually has today.

How to check a specific rollup’s sequencer setup

Three practical questions cut through most of the marketing language: who currently runs the sequencer, is there a working forced-inclusion path to L1 that bypasses it, and what happens to users’ funds and ability to transact if that sequencer stops responding. A rollup’s own documentation and its security-model disclosures, cross-checked against independent trackers of L2 security properties, are more reliable answers than a roadmap announcement.

Why single sequencers became the default

Running a sequencer well requires low-latency infrastructure, careful handling of transaction ordering, and tight integration with whatever proving or batching pipeline the rollup uses. Building that as a single, centrally operated service was simply the fastest way to ship a working rollup, and it let teams iterate quickly on execution-layer features without also having to solve decentralized ordering at the same time. That sequencing choice was a deliberate, acknowledged shortcut from the start, not an oversight — most rollup teams that run a single sequencer today say so plainly in their own documentation, and most describe decentralizing it as a roadmap item rather than a solved problem.

What users actually experience day to day

For the overwhelming majority of transactions, a centralized sequencer is invisible: it orders transactions roughly on a first-come basis, includes them promptly, and users never need to think about who is doing that work. The risk is not that every transaction is at risk of censorship — it is that the capability to censor, reorder for profit, or simply go dark exists concentrated in one place, and that the rollup’s design determines how much that capability is checked by anything else. A well-designed forced-inclusion path converts “the sequencer could censor me with no recourse” into “the sequencer could censor me, but I have a slower, costlier way around it” — a real improvement, but not the same as removing the central point of control.

Reading a rollup’s own disclosures

Rollup teams increasingly publish their own security and decentralization status, sometimes self-scored against frameworks like the one described in the companion guide on L2 security models. Those disclosures are a reasonable starting point, but they describe the project’s own account of itself; cross-referencing against independent trackers of L2 properties, and checking whether a forced inclusion path has actually been used or merely specified, gives a more complete picture than the roadmap slide alone.

It is also worth separating two different complaints people make about sequencers. One is about decentralization of ordering power itself; the other is about the fairness of the ordering a sequencer applies, which touches directly on MEV. A rollup can decentralize who runs the sequencer without necessarily changing how MEV is distributed among the parties involved, and vice versa — the two problems are related but not the same problem wearing different names.

A practical habit worth building: before moving meaningful value through a rollup, check whether its sequencer has a documented outage history and whether its forced-inclusion path has ever actually been exercised in production rather than only described in a whitepaper. A mechanism that has never been used under real conditions is a weaker assurance than one that has already been tested by an actual outage.

Treat sequencer centralization as an operational risk disclosure, not a reason by itself to avoid a rollup — most of the ecosystem shares this limitation today, and the relevant question is how each project has mitigated it, not whether the limitation exists at all.

Risk note

A centralized sequencer is an operational and censorship risk, not evidence that a rollup’s posted state is fraudulent. The two are separate failure modes: one is about who controls ordering today, the other is about whether the proof system backing settled state is sound.

Not financial advice. Nothing on this page is a recommendation to buy, sell, stake, bridge, or avoid any asset or protocol.

FAQ

If the sequencer goes down, do I lose my funds?

No. A sequencer outage halts new transaction processing on that rollup; it does not affect state that has already settled on L1. New activity resumes once the sequencer or a fallback mechanism comes back online.

What is forced inclusion?

A mechanism, implemented directly on L1, that lets a user submit a transaction even if the rollup’s sequencer refuses to include it. It is typically slower and costlier than the normal path, and its exact availability varies by rollup.

Are shared sequencers live today?

Shared sequencer networks exist at varying stages of development and adoption across the ecosystem; check a specific project’s own documentation for its current status rather than assuming it is live everywhere.

Does a decentralized sequencer fix MEV?

It changes who can extract MEV and how competitive that process is, but it does not eliminate MEV as a phenomenon. See the companion guide on MEV in plain English for how ordering value is captured more broadly.

Frequently asked questions

If the sequencer goes down, do I lose my funds?
No. A sequencer outage halts new transaction processing on that rollup; it does not affect state that has already settled on L1. New activity resumes once the sequencer or a fallback mechanism comes back online.
What is forced inclusion?
A mechanism, implemented directly on L1, that lets a user submit a transaction even if the rollup's sequencer refuses to include it. It is typically slower and costlier than the normal path, and its exact availability varies by rollup.
Are shared sequencers live today?
Shared sequencer networks exist at varying stages of development and adoption across the ecosystem; check a specific project's own documentation for its current status rather than assuming it is live everywhere.
Does a decentralized sequencer fix MEV?
It changes who can extract MEV and how competitive that process is, but it does not eliminate MEV as a phenomenon. See the companion guide on MEV in plain English for how ordering value is captured more broadly.

Educational content, human-edited and possibly AI-assisted. It describes how systems work and is not financial, legal or tax advice. Report an error via our corrections page.

More in Rollups

Related articles

See all