Base post-mortem reveals sequencer bug behind back-to-back outages
Base suffered two mainnet outages last week because of a sequencer bug in block-building logic. An invalid transaction failed as expected, but the sequencer did not clear stale journal state, leaving accessed accounts and storage slots cached incorrectly. Since Base uses a single sequencer, the bug halted block production entirely. The first outage lasted 116 minutes and the second 20 minutes. A patch fixed the journal-state handling, but recovery was delayed by unrelated infrastructure issues, and a race condition after reset caused the second outage. Planned improvements include stronger fuzz testing and more graceful recovery so validator nodes can restart without manual intervention. Base has had prior sequencer-related outages, including brief incidents in 2024 and 2025.
