We didn’t ask the right question until the emergency multisig failed.
It was 3 AM in Tallinn, and I was staring at a block explorer for a Layer2 that had just processed its millionth transaction. The network was glowing — total value locked above $2 billion, gas fees a fraction of Ethereum mainnet, and a chorus of tweets celebrating "the future of scaling." But one line in the explorer caught my eye: the sequencer address was controlled by a single EOA. One key. One person could stop the chain.
I felt a familiar rush — the kind you get when you realize the story everyone is telling doesn’t match the code. This wasn’t a rogue project; it was the most popular rollup in the market. The team had promised "progressive decentralization" in their whitepaper. A year later, the sequencer was still a single node running on AWS. We didn’t call it what it was — centralized infrastructure with a decentralized marketing budget.
Context: The Layer2 Promise vs. The Layer2 Reality
Layer2 scaling solutions emerged from a simple necessity: Ethereum’s base layer could not handle global adoption. Rollups — both optimistic and zero-knowledge — offered a path to 1,000+ transactions per second without sacrificing security. The narrative was seductive: move execution off-chain, keep settlement security on mainnet, and eventually hand over control to a decentralized set of sequencers. That "eventually" has become the industry’s longest-running PowerPoint slide.
I’ve been tracking Layer2 architectures since 2021, when the first Optimism mainnet launched. At that time, the sequencer was centralized by design — a "training wheels" phase. The roadmap promised a phased upgrade to decentralized sequencing, with multiple operators, threshold signatures, and economic slashing. Three years later, most rollups still run a single sequencer. A few have added a second or third operator, but that’s not decentralization — it’s a federation of insiders.
The problem isn’t technical incompetence. Multiple teams have built functional decentralized sequencer prototypes. Espresso, Astria, and others have testnets that rotate proposers every few blocks. Yet production rollups have not adopted them. Why? Because decentralized sequencing introduces latency, complexity, and — most importantly — a loss of control for the founding team. The sequencer is the most profitable component: it captures MEV, orders transactions, and can front-run or censor. Handing that to a permissionless set means giving up the business model.
Core: The Technical Anatomy of a Centralized Sequencer
Let’s walk through what a centralized sequencer actually does in a typical optimistic rollup. The sequencer receives transactions, orders them, creates a batch, and submits that batch to L1. It also provides fast confirmation — users see finality in seconds, not minutes. This is the core value proposition. But the sequencer also holds the power to reorder transactions, exclude them, or even halt the chain.
During my work on a sovereign rollup project in 2023, I audited the sequencer module of a popular rollup SDK. The default implementation had a single private key signing all batch submissions. There was no mechanism for rotation, no governance for key replacement. The team argued that "we’ll decentralize later" — but the codebase had no hooks for it. The entire architecture was built around a trusted operator. Adding decentralization later would require a complete rewrite of the sequencer logic.
This isn’t an isolated case. I’ve reviewed five separate rollup codebases over the past two years. All of them hardcoded the sequencer address in the genesis configuration. None had a built-in fallback mechanism if the sequencer went offline. The only redundancy was a backup server in a different AWS region. That’s not blockchain — that’s a cloud service.
— Root: The architectural debt is so deep that even if teams wanted to decentralize now, it would take six to twelve months of engineering effort. Most don’t have the incentive. The market rewards speed and user experience, not sovereignty. Users love the fast confirmations and low fees. They rarely check whether the sequencer is a single point of failure. The few who do are dismissed as "purists."
But the risk is real. If a centralized sequencer is compromised, the attacker can produce invalid state roots, drain bridges, or censor transactions for weeks before L1 fraud proofs catch up. The optimistic rollup’s fraud proof window is typically 7 days. That’s a week where the sequencer has absolute power. In a bear market, the cost of attacking a sequencer is low — the operator’s reputation is the only collateral. That’s not a security model; it’s an honor system.
Contrarian: Maybe Centralized Sequencing Is Good Enough — For Now
I’ve spent years arguing for decentralized sequencing. But recently, I’ve started to question whether the obsession is misplaced. Consider the alternative: a permissionless sequencer set with 21 nodes, each requiring a large stake. Who can afford that? The same venture capital firms who already control the ecosystem. Decentralization doesn’t automatically mean distribution of power — it often means a cartel of wealthy validators.
Moreover, the user experience of a decentralized sequencer is objectively worse. Latency increases, reorgs become more common, and the economic cost of running a node rises. Coinbase’s Base chain thrives on a single sequencer because it’s fast and cheap. Users don’t care about the sequencer’s decentralization — they care about the bridge being secure (which is secured by Ethereum L1) and the UX being smooth.
— Root: The market is voting with its transactions. The most successful Layer2s are the most centralized. Arbitrum and Optimism have tokenized governance but still operate centralized sequencers. The "decentralization" narrative is a regulatory and marketing shield, not a technical necessity. Perhaps the real trade-off isn’t centralization vs. freedom, but performance vs. resilience. In a bull market, performance wins every time.
But here’s the blind spot this argument misses: bull markets don’t last. When the next crash comes, centralized sequencers become attack vectors. We’ve seen it before — the Solana outage cascade was partly due to centralized sequencer-like mechanisms. A determined attacker can target the sequencer, cause chaos, and drain bridges. The lack of decentralization becomes a systemic risk that only surfaces during stress.
I remember a conversation with a rollup founder in Lisbon, 2024. He admitted that his team had no plan to decentralize the sequencer "until a competitor does it first." That’s the prisoner’s dilemma of Layer2. No single player wants to move first because it increases cost and reduces revenue. So the industry stays stuck in a suboptimal equilibrium — all marketing, no substance.
Takeaway: The Fork That Should Have Happened
We didn’t push hard enough when it mattered. In 2022, the Ethereum community debated mandatory sequencer decentralization for rollups. The idea was to include it in the rollup standardization frameworks. But the pushback was fierce: "let the market decide," "don’t force premature decentralization." The market decided — it chose speed over sovereignty. Now we have a multi-billion dollar ecosystem built on trust in a handful of sequencer operators.
I’m not advocating for immediate, forced decentralization. That would break the user experience and likely kill several projects. But we need to stop pretending that today’s Layer2s are decentralized. They are remote procedure call providers with a blockchain wrapper. The real innovation — fully trustless, decentralized sequencing — is still a research problem. And the industry has shown little appetite to solve it.
The next bull run will likely mask these issues again. TVL will surge, fees will rise, and everyone will celebrate the "scaling solution." But underneath, the sequencer will still be a single point of failure. Until that changes, the word "Layer2" is a misnomer. It’s not a second layer of trust minimization — it’s a second layer of convenience.
So the question I leave you with is not "when will sequencing be decentralized?" but "who will pay the cost when it isn’t?" The answer, as always, is the user — the last one to find out and the first one to lose their funds.
— Root: The next time you see a rollup celebrate its "secuarity" or "decentralization," look at the sequencer address. If it’s a single EOA, you’re not using a blockchain. You’re renting a server.
