Coinbase’s Base network went down twice in the span of a single week, and a freshly published post-mortem explains exactly why.
The culprit is a bug in the sequencer’s block-building logic. When an invalid transaction came in, the system rejected it — business as usual. But it failed to clear the journal state, which kept stale data about accounts and storage slots floating around in memory. That orphaned state meant the sequencer couldn’t pick up where it left off.
Base’s architecture runs on a single sequencer, so one hiccup can grind everything to a halt. The first outage lasted 116 minutes. The second, just 20 minutes — caused by a race condition after a system reset that stopped sequencers from catching up.
The patch was straightforward: force the journal state to properly update during execution. Recovery took longer than the team expected, though, thanks to infrastructure issues unrelated to the bug itself.
Base has dealt with sequencer trouble before. There was a 17-minute blockage back in September 2024 and another roughly 30-minute outage in August 2025. The team says two planned improvements should help going forward: more aggressive fuzz testing of the protocol and graceful recovery mechanisms so validator nodes can bounce back without manual restarts.
To put this in perspective, Base is the second-largest layer-2 by total value locked, sitting just under $11 billion.
It’s worth noting that relying on a single sequencer is still the weak link here. Other chains like Arbitrum, OP Mainnet, and zkSync Era have faced similar single-point-of-failure outages. The engineering tradeoffs are real, but so are the consequences when things break.
