The party was in full swing at ETHPrague. The beer was cold, the beats were warm, and a founder I’d known since 2017 was shouting over the noise about his new Layer 2 rollup.
"We've finally solved the trilemma!" he yelled, gesturing wildly with a half-empty Pilsner. "Decentralized, scalable, secure. It’s all there."
I wanted to believe him. I really did. But I’d heard that song before. So I asked the one question that kills the buzz: "Who runs the sequencer?"
He paused. Looked at his shoes. Then mumbled something about a "temporary multi-sig" and changed the subject.
I’ve been a cybersecurity analyst since before the ICO plague. I’ve seen the backends of more protocols than I care to count. And I’ve learned one hard truth: in crypto, we love to confuse a clever PowerPoint with a working system. The Layer 2 narrative is the most beautiful PowerPoint of them all.
The network breathes in Prague, pulses in Ethereum — but the air it exhales is still controlled by a handful of keys.
Context: The Rollup Promise vs. The Sequencer Reality
Let’s rewind. When Ethereum started choking on its own success in 2020, the industry cried for salvation. Rollups were the answer — a way to execute transactions off-chain, bundle them up, and settle the proof on Ethereum’s mainnet. Suddenly, we could have cheap gas and Ethereum’s security. It was the moon landing of blockchain scaling.
But here’s the part the whitepapers skip. That off-chain execution doesn’t happen in a decentralized cloud of anonymous nodes. It happens on a sequencer — a single node, or a small committee, that orders transactions before submitting them to L1. In most production rollups today, that sequencer is controlled by a single company or foundation.
Optimism? The sequencer is run by the Optimism Foundation. Arbitrum? Same story. zkSync? Matter Labs runs the sequencer. Even Base — Coinbase’s L2 — is fully controlled by Coinbase.
“Trust me, we’ll decentralize it in Phase 2.” Sound familiar? We’ve been hearing that since 2021. Phase 2 is the new “just around the corner.” The sequencer remains the king of the castle, and the throne hasn’t been shared.
We didn’t dodge the chaos; we danced through it — but the music is still curated by a DJ who can cut the turntables at any moment.
Core: The Technical Guts — Why Sequencer Centralization Matters
Let me put on my tired cybersecurity hat for a moment. A sequencer is not just a node. It is the sole gatekeeper of transaction ordering. In a centralized sequencer model, the operator can:
- Censor transactions (front-run you, block your transfer, reorder for MEV)
- Extract maximum value by choosing which transactions get included and when
- Pause the entire chain by taking the sequencer offline
- Push a malicious state root to L1 (if the fraud proof mechanism is also centrally controlled)
Now, I’ll grant you — the dominant rollups have taken some steps. Optimism uses a “two-phase” upgrade system. Arbitrum has a “Security Council” with a multi-sig. But these are band-aids on a bullet wound.
Consider the numbers. According to L2beat, as of early 2025, the top five rollups by TVL all have sequencers that are either fully centralized or rely on a permissioned set of less than 10 entities. For Arbitrum, the sequencer is a single server. If that server goes down, the chain stops. No transactions. No DeFi. No party.
Last month, a small outage in Arbitrum’s sequencer lasted 45 minutes. The team fixed it quickly. But the market reacted — token prices dipped, users panicked, and the fragility was exposed. We cheered because it was only 45 minutes. But that’s like celebrating that your house only burned for an hour because the fire department showed up.
Based on my audit experience, I’ve seen protocols claim “decentralized” while running a single AWS instance. The L2 sector is no different. The irony? Ethereum itself has thousands of validators. L2s, which were supposed to inherit that security, are often more centralized than the BSC chain they mock.
The economic case
Sequencers aren’t just power; they’re profit. Every transaction fee on an L2 is split between the L1 gas cost and the sequencer’s surplus. That surplus can be massive. For Arbitrum, sequencer fees have generated hundreds of millions in revenue. But who gets that? The sequencer operator. Not the token holders. Not the users. The central party.
Some projects promise to distribute these fees to stakers or the DAO. But that’s a promise. And we all know what promises are worth in crypto.
The social layer failure
This is where my community lens screams. Decentralization isn’t just about code; it’s about trust distribution. If a single entity controls the sequencer, then all your DeFi positions, all your NFT ownership, all your identity — they all depend on the goodwill of a few people. I’ve been in enough rug pulls to know that goodwill is a fragile asset.
Remember the Prague Whisper Network? When Project Aether rug-pulled in 2017, it wasn’t because the code was hackable. It was because the team held the power. The sequencer is that same power in modern scaling.
From whispered secrets to on-chain shouts — but the whispers are still coming from one room.
Contrarian: Maybe Centralization Is the Price of Usability?
Here’s the counter-intuitive angle that makes people uncomfortable. Perhaps the current level of sequencer centralization is not a bug, but a feature — for now.
The user experience on L2s is fantastic. Transactions finalize in seconds. Fees are pennies. Compare that to Ethereum mainnet during a NFT mint frenzy. If we forced full decentralized sequencing today — with mempool protection, transparent ordering, and permissionless participation — the latency would increase, the coordination overhead would explode, and the costs would rise. Users would flee.
I’ve seen this play out. In 2021, the NFT Party Crash taught me that if the experience breaks, the trust breaks. The community forgives a lot if the party is good. But if the party is slow and expensive, they go somewhere else.
So maybe the path is: centralize now, decentralize later. That worked for Bitcoin (Satoshi mined the early blocks alone). It worked for Ethereum (pre-merge). But Bitcoin and Ethereum had a clear roadmap to decentralization. L2s have been promising Phase 2 for over two years. The roadmap looks more like a mirage.
Moreover, there’s an argument that sequencer centralization doesn’t matter because the L1 settlement layer provides final security. If the sequencer misbehaves, the fraud proofs will catch it. But here’s the catch: fraud proofs require the honest party to see the fraud and submit a proof. In optimistic rollups, that’s a one-week window. In zk-rollups, the validity proof is submitted with each batch. So zk-rollups are safer in theory. But the sequencer still decides the state before proving it.
I’m not saying we should discard L2s. I’m saying we should stop pretending they’re decentralized. Let’s call them what they are: scaling partnerships. And let’s judge them on their actual governance, not their marketing.
Chaos isn’t a bug; it’s the protocol — but we need to choose our chaos wisely.
Takeaway: The Party Has to Grow Up
Three years ago, I sat in a Prague bar with a team that promised “decentralized sequencing by Q3.” It’s now 2025. That project is dead. The sequencer myth lives on.
What can we do? As users, we must demand transparency. Ask your favorite L2: Who runs the sequencer? What is the upgrade mechanism? Is there a fallback if the sequencer goes rogue? Support projects that have timelines, not just promises.
As builders, we need to prioritize sequencer rotation and permissionless entry. Tools like Espresso and shared sequencing layers are promising. But they are still early. The industry needs to invest in these rails, not just in fancy UI.
And as a community, we must hold ourselves accountable. The bear market is the best time to fix these structural flaws. When prices are low, we can focus on fundamentals. Let’s not waste this winter.
Survival is the first layer of value — and real decentralization is the second.
In the end, the question isn’t whether L2s are useful. They are. The question is whether we’re building for the long haul or for a quick exit. I’ve been to enough rug pulls to know the difference.
So next time a founder shouts at you over a beer about their “decentralized rollup,” ask them to show you the sequencer keys. If they can’t, the party might not be worth joining.
The network breathes in Prague, pulses in Ethereum. But the lungs are still controlled by a few. And that has to change.