Base Pinpoints Persistent Sequencer Bug Behind Back-to-Back Outages

The root cause originated from an invalid transaction that failed during execution. Instead of clearing, the system retained a stale journal state within the block builder, inadvertently caching account and storage data from the failed attempt. When the network processed a subsequent valid transaction, it applied this corrupted state, leading to an invalid state transition that forced nodes to reject the block.

This technical deadlock caused transaction queues to overflow, resulting in rejected requests from users. Base engineers eventually deployed a patch to ensure the journal state updates correctly after transaction failures. However, recovery proved difficult due to a secondary race condition in the engine reset feature, which hindered sequencers from synchronizing properly after the initial restart.

To prevent further disruption, the development team is now overhauling its testing infrastructure. Plans include more rigorous fuzz and load testing to identify anomalous transaction patterns before they hit the mainnet. Additionally, Base is introducing graceful recovery protocols to help validator nodes maintain synchronization during future engine failures. These updates follow the recent deployment of the Beryl upgrade, which introduced the B20 token standard and shortened withdrawal periods from seven days to five.