On June 25 and 26, 2026, Base mainnet, an L2 built on the OP Stack, suffered two block production outages lasting approximately 116 minutes and 20 minutes respectively. On June 28, Base published a detailed postmortem confirming that no on-chain assets were compromised and user funds remained accessible throughout the incidents.
Event Overview
The first outage occurred around 13:30 UTC on June 25, disrupting block production for nearly two hours. The second outage occurred at approximately 14:10 UTC on June 26, lasting about 20 minutes. Both events were network-wide failures that prevented users from submitting transactions. Base engineering teams quickly identified the root causes and deployed fixes.
Root Cause: Sequencer State Management Bug
The postmortem identified a critical defect in the sequencer's block-building logic. During normal operation, the sequencer maintains a journal of intermediate state for each transaction in a block. If a transaction fails (e.g., due to insufficient gas or invalid signature), the journal should be rolled back to a clean state before processing the next transaction. However, in this incident, the system failed to clear the historical journal after a failed execution. As a result, subsequent valid transactions incorrectly reused the gas calculation context from the failed transaction, leading to the generation of an invalid state transition block. Once broadcasted, this invalid block broke the consensus of the entire L2 network, causing all sequencers to halt block production.
The bug was subtle because it only manifested under specific conditions: a failed transaction followed by a valid transaction that would normally succeed. Under heavy mempool pressure, such sequences became more likely, explaining why the outage occurred during normal network activity rather than during a targeted attack.
Network Impact and User Experience
During the two outages, Base mainnet exhibited a complete cessation of block production. No new blocks were added to the chain, and pending transactions could not be included. The mempool became congested as users retried failed submissions. Developers interacting with the network via eth_sendRawTransaction received persistent error responses. Despite these disruptions, Base emphasized that all previously confirmed on-chain assets remained safe and fully available; the outage did not affect the integrity of the L2 state root committed to L1, nor did it compromise any user funds.
Fix and Second Outage Cause
Base engineers developed patch PR #3806 to correct the journal state cleanup during block building. After applying the patch and restarting the sequencer cluster, block production resumed successfully. However, the restart process itself introduced a race condition in the engine initialization: certain sequencer nodes reset their internal engine state before fully synchronizing with peers, causing a temporary mismatch in the pending block list. This led to a second, shorter outage on June 26 while the cluster re-synchronized. Base has since applied a separate fix to synchronize the engine reset sequence across all nodes, preventing the race condition from recurring.
Future Improvements
To strengthen network resilience and minimize future downtime, Base announced three key initiatives. First, they will enhance protocol-level fuzz testing and stress testing capabilities. By systematically generating a wide variety of transaction sequences (including edge cases like failed-then-valid patterns), Base aims to catch similar state management defects before they reach production. Second, they plan to upgrade the monitoring and operations stack with real-time anomaly detection alerts for block production stalls and sequencer state inconsistencies. Third, Base will implement a graceful recovery mechanism: instead of requiring manual intervention to restart services after a failure, the network will automatically detect an unhealthy state, roll back to the last valid state, and resume block production within minutes. This mechanism is being designed to handle both journal-related bugs and unexpected restart conditions.
These improvements reflect a broader industry trend toward proactive reliability engineering in L2 networks, where even brief outages can erode user trust. Base’s transparent postmortem and concrete roadmap serve as a model for how L2 teams should respond to production incidents.