Should Sequencers Be Decentralized? Layer2 Performance and Malicious Behavior Issues
TechFlow Selected TechFlow Selected
Should Sequencers Be Decentralized? Layer2 Performance and Malicious Behavior Issues
What's so impressive about Layer2?
Navigating Web3 tides with focused insights
Author: Lao Bai, Blockchain Analyst at Amber Group
What Makes Layer2 So Powerful?
The most impressive aspect of Rollup is this: while nearly (but not 100%) inheriting the security of L1, it can push performance to the extreme.
Why?
Because blockchains are slow precisely due to the requirement that multiple nodes must process and propagate all transactions, and eventually reach consensus.
If ETH had only one node, its TPS should be several thousand at minimum. But with so many decentralized nodes, TPS drops to just over ten.
Therefore, the main directions for scaling various Alt Layer1s are essentially these:
-
Improving node performance——processing more transactions per unit time;
-
Reducing the number of nodes——decreasing communication among nodes and shortening consensus time;
-
Optimizing propagation mechanisms——improving bandwidth utilization and reducing the proportion of bandwidth consumed by consensus protocols;
-
Not requiring every node to process everything——similar to an assembly line in a factory, division of labor increases efficiency (sharding, layering, modularity... Flow being an exception).
From early projects like EOS to recent ones like Aptos, all Alt Layer1s focus on one or two of these four broad directions, combined with minor technical innovations—just different approaches and trade-offs.
Why Can Layer2 Push Performance to the Extreme?
Because theoretically, it can run on just one node. It can infinitely scale factor 1, while completely ignoring factors 2, 3, and 4.
So what if "I" am just a centralized official Sequencer? As long as "I" have OP or ZK proofs to guarantee the safety of your funds, who cares?
Thus, there's no need for consensus, no need for complex propagation mechanisms, no need for分工合作—essentially identical to the client-server model in Web2.
Where Is the Performance (TPS) Limit of Layer2?
It will most likely be determined by the lower of the following two values:
-
The maximum processing capacity of a single Sequencer;
-
The gas limit for blocks on Ethereum L1 used as the settlement layer (currently around 30 million).
Should Layer2 Sequencers Be Decentralized?
To answer this question, we first need to ask: what are the advantages and disadvantages of *not* decentralizing?
-
Advantage:Can achieve maximum TPS. Decentralizing across multiple Sequencers inevitably introduces consensus and data propagation overhead, leading to performance loss.
-
Disadvantage:Less ideologically "blockchain-like," and potential risks of malicious behavior by the Sequencer.
Can a Sequencer Act Maliciously?
The essence of Rollup lies in using Optimistic Proofs (OP) or Zero-Knowledge (ZK) proofs to allow Layer2 to inherit Layer1’s security.
However, this security is theoretically only infinitely close—not perfectly equivalent.
Therefore, a Sequencer does have a small window for malicious behavior, especially when it's a single centralized entity.
But where exactly?
It's not about altering transactions or fabricating fake ones—because the Sequencer must periodically submit raw transaction data to Layer1, and full nodes elsewhere download this data from Layer1 to monitor the Sequencer.
Due to digital signatures, the Sequencer cannot submit forged or modified "invalid transactions"—even if it did, such submissions would be meaningless.
There Are Only Two Things a Sequencer Can Do
-
Hide legitimate transactions——equivalent to censorship;
-
Reorder transactions——the Sequencer could maintain two sets of "valid ledgers," and other full nodes might not detect the discrepancy.
An example of hiding legitimate transactions: A and B conduct an OTC trade where A sends ETH to B. Two seconds later, B sees the receipt and transfers money to A's bank account. However, A has bribed the Sequencer, which simply never submits A’s ETH transfer to B onto L1. Minutes later (Arbitrum submits data to Layer1 every three minutes; OP takes even longer), the ETH in B’s wallet disappears out of nowhere.
A dual-ledger scenario could look like this: During a highly anticipated NFT mint, someone colludes with the Sequencer to monopolize mint slots. The Sequencer cannot openly reject other users’ mints, so it appears to respond normally to requests, but secretly maintains a forked ledger where all NFTs are minted first by the briber. Eventually, only this forked ledger is submitted to Layer1, causing many users who thought their mints succeeded to find them invalidated after settlement.
Ultimately, the result given within two or three seconds on Layer2 is only Layer2 finality—not finality on the Layer1 settlement layer. True finality occurs only after Rollup data has been fully submitted and confirmed on Layer1 (though technically, Layer1 itself has a ~15-minute finalization period; Vitalik is pushing to reduce this to single-slot finality in 12 seconds—a topic beyond the scope of this article). This few-minute window is the theoretical "vulnerability period" during which a Sequencer could act maliciously.
So, should Sequencers be decentralized?
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
Lao Bai
