Understanding Sequencer Decentralization: A New Framework for Evaluating L2 Rollups

Intro

‍

“Decentralized sequencing” currently refers to different things across the ecosystem, ranging from a small set of foundation-operated nodes to fully permissionless networks. Without a clear framework to distinguish between these designs, the ecosystem risks making important decisions based on claims rather than architecture.

‍

Today, Nethermind Research is releasing a comprehensive framework (developed in collaboration with L2BEAT) that establishes clear criteria so builders, users, and institutions can evaluate sequencer designs based on what they actually deliver.

‍

‍

‍

Why Sequencer Decentralization Matters

‍

Most rollups today rely on centralized sequencers. These sequencers are single entities responsible for ordering transactions, submitting state updates to Ethereum, and maintaining rollups' day-to-day operations. While user funds remain secure through Layer 1 enforcement, centralized sequencers can:

‍

• censor transactions for extended periods
• reorder them for MEV extraction, or
• go offline and freeze the rollup.

‍

Escape hatches exist, but they are costly safeguards rather than practical alternatives for normal use.

There is also a regulatory dimension. SEC Commissioner Hester Peirce has warned that rollups with centralized sequencers may be treated as securities exchanges, which could introduce registration requirements and limit permissionless access.

‍

Centralization is not only a technical choice. It affects the long-term viability of rollups as open, neutral infrastructure.

‍

‍

‍

Understanding Different Approaches to Sequencing

‍

Projects use the term “decentralized sequencing” to describe a wide range of designs. Some begin with a small set of operators, while others use permissionless networks or rotating committees. These approaches sit at different points on the decentralization spectrum, but without a shared framework, they are difficult to compare.

Each approach introduces different trust assumptions. The Layer 1 protects user funds, but properties like liveness, censorship resistance, and ordering fairness depend on the sequencing architecture itself. A decentralization roadmap can outline intent, but the guarantees available today come from the design that is currently deployed.

‍

‍

‍

The Five Properties of Decentralized Sequencing

‍

A well-designed decentralized sequencer should provide:

‍

1. Preconfirmation Safety distributes trust across multiple sequencers instead of concentrating it in one, improving the reliability of soft finality for the rollup's transactions.
2. Practical Liveness ensures the rollup continues operating even if individual sequencers fail, without requiring users to fall back to Layer 1.
3. Censorship Resistance makes censoring transactions significantly harder by requiring coordination among many sequencers.
4. Fair Ordering reduces monopolistic control over transaction sequencing and opens the door to ordering schemes that limit MEV manipulation.
5. Credible Neutrality establishes rollups as trustworthy public infrastructure rather than centrally managed platforms.

‍

→ Read the full technical paper for definitions, scoring rules, and methodology.

‍

‍

‍

The Risk Framework

‍

Our framework evaluates sequencer decentralization across four critical dimensions, with each scored as Low Risk (🟢), Medium Risk (🟡), or High Risk (🔴):

‍

• ‍Preconfirmation Safety measures how reliably soft-finalized transactions can be trusted. Economic security is evaluated by comparing Total Slashable Stake (TSS) to Total Value Secured (TVS). TSS is considered robust when it exceeds one-third of TVS.‍
• Practical Liveness assesses whether the rollup continues operating if sequencers fail. Low risk systems rely on Byzantine Fault Tolerant protocols with automatic recovery.‍
• Censorship Resistance evaluates how difficult it is to censor transactions. The strongest designs enforce inclusion rules at the protocol level.‍
• Fair Ordering examines whether transaction ordering follows transparent rules or is left to proposer discretion.‍
• Credible Neutrality assesses whether sequencing avoids structural advantages for specific actors and aligns with open, public infrastructure principles.

‍

‍

‍

What the Framework Reveals

‍

We applied the framework to several prominent rollup designs to show how different architectural choices shape sequencing decentralization. Each ecosystem emphasizes certain properties while accepting trade-offs in others.

The goal is not to rank projects. It is to provide a consistent way to understand how sequencing designs differ. Some systems prioritize credible neutrality by aligning closely with Ethereum Layer 1. Others emphasize permissionless participation or focus on high-performance protocols that offer fast confirmations. Hybrid approaches combine elements of both, such as using based sequencing with optional preconfirmations.

‍

The full technical report includes project-by-project evaluations for Starknet, Aztec, Taiko, and Surge, along with an analysis of their architectural choices and decentralization roadmaps.

‍

→ Explore the project evaluations in the full report.

‍

‍

‍

Key Design Components

‍

Three design choices determine how decentralized a sequencer network can be.

‍

1. ‍Network Composition distinguishes between systems that are fully permissionless and those that restrict participation through whitelisted entities. Committee-based selection from permissionless sets can still trend toward centralization when committees are small, or selection is heavily stake-weighted.‍
2. Consensus Mechanisms vary in resilience. Classical BFT protocols that tolerate up to one-third Byzantine faults offer strong liveness through automatic recovery. Optimistic consensus mechanisms reduce latency but offer weaker fault tolerance, which can be risky for small operator sets.‍
3. Proposer Selection shapes how sequencing power is distributed. Deterministic rotation creates predictable schedules that can be attacked. Randomized selection reduces predictability but does not eliminate concentration when stake dominates selection probability. Auction-based mechanisms often concentrate power among sophisticated actors.

‍

These components cannot be evaluated in isolation. A permissionless network with weak fault tolerance may offer fewer guarantees than a smaller committee with robust BFT recovery.

‍

‍

‍

What This Means

‍

Different teams will prioritize decentralization properties differently. This framework provides builders, developers, and stakeholders with a clearer understanding of those trade-offs. It enables them to evaluate where a design stands today in relation to its longer-term roadmap.

‍

A key insight from the report is how economic security shapes the trustworthiness of preconfirmations. The framework measures this by comparing the Total Slashable Stake (TSS) to the Total Value Secured (TVS) in the rollup.

‍

Across the examples we studied, a clear pattern emerges:

‍

• Below one-third TSS/TVS, soft-finality guarantees are weak.
• Above one-third, attacks become economically costly, making preconfirmations far more reliable.

‍

This gives teams a practical benchmark for assessing whether a sequencer set is secure today, not just in theory.

‍

The full report offers a more in-depth examination of this methodology, including case studies that demonstrate how different rollups perform under the TSS–TVS model. We invite the community to explore the details and apply the framework to emerging sequencer designs.

‍

‍

‍

Conclusion

‍

Decentralized sequencing is not a single model. It spans a range of architectures, each with distinct strengths and trade-offs. This framework provides a structured approach to evaluating those differences and bringing clarity to an area where terminology often obscures meaningful differences. As rollups continue to evolve, we hope this contributes to clearer standards and more robust designs for decentralized sequencing.

‍

This framework is a collaboration between Nethermind Research and L2BEAT, building on L2BEAT’s established risk assessment methodologies including the Stages Framework, the Risk Rosette, and domain-specific frameworks for data availability and trusted setups.