The Role of the Sequencer

A rollup needs someone to receive transactions, order them and produce blocks. On almost every rollup in production, that someone is a single operator.

What the sequencer does

Accepts transactions from users, orders them, executes them to produce a new state, and periodically posts the data and the resulting state to the base layer.

Because it is one party, ordering is immediate and there is no consensus delay. That is why rollup confirmations feel instant compared with base layer blocks.

What it can do

Order transactions however it likes. Including extracting value from that ordering, or placing its own transactions favourably.

Refuse to include a transaction. Censorship, temporarily.

Stop. In which case no new transactions are processed by the normal route.

What it cannot do

Steal funds. The state transitions it posts must be valid, enforced either by fraud proofs or by validity proofs. An invalid state is rejected regardless of who proposed it.

Prevent withdrawal permanently. Every credible rollup has a mechanism for submitting transactions directly to the base layer, bypassing the sequencer entirely.

That escape hatch is the reason sequencer centralisation is tolerable. It is slow and expensive, and it exists.

The escape hatch in practice

Forced inclusion works by submitting your transaction to a contract on the base layer. The rollup is required to include it within a defined window, or the rollup itself is considered to be misbehaving.

The mechanism is rarely used and its existence constrains the sequencer’s behaviour, because censorship that can be routed around is not censorship.

Whether it has been tested under stress varies by rollup, and independent trackers publish assessments of each network’s arrangements.

What happens when the sequencer stops

Transactions stop being processed through the normal route. Balances are unaffected.

Users can still exit through forced inclusion, at base layer cost and on base layer timescales.

Outages have occurred on major rollups, typically lasting from minutes to hours.

The direction

Shared and decentralised sequencing is under active development, with several designs deployed or being tested. The objective is to remove the single operator without losing the latency advantage.

Progress is real and it is not complete anywhere, which is why the honest description of any rollup today includes a single sequencer and a documented escape hatch.

For a user

The practical consequence is that funds on a rollup are subject to an operator’s availability for the fast path, and recoverable slowly regardless.

For anything where that matters, routing through a venue that supports the rollup, such as exchanges that let you specify the network, gives an exit that does not depend on the sequencer being available.