Sui mainnet restored operations after a six-hour outage

The Sui team emphasized that the outage was not caused by overload, heightened activity, or an external attack. The root issue was a rare bug in the consensus commit logic triggered by a specific data-cleaning condition. It caused some validators to compute different consensus outputs and generate mismatched checkpoint candidates.

Because more than one-third of the stake signed different digests, the required quorum could not be met. The network deliberately paused progress to avoid finalizing an inconsistent state.

Developers noted that this is the failure mode Sui’s architecture is designed to handle: at any sign of divergence, nodes halt advancement to ensure no forks or transaction rollbacks occur.

During the outage:

• transactions did not execute;
• submissions ended in timeouts;
• RPC reads continued to return the last certified state;
• user funds and data remained fully secure.

Recovery proceeded in three steps. The team first identified the cause of the divergence and prepared a fix to remove incorrect consensus data and resynchronize logic. The update was then deployed on Mysten Labs validators in a test environment to confirm correct behavior. Afterward, most validators rapidly applied the patch, re-ran consensus, and resumed signing checkpoints.

The network returned to normal once quorum was reached on the next checkpoint.

The Sui team said it plans to strengthen safeguards by improving early detection of inconsistent states, automating operator tools, and expanding testing of consensus-failure scenarios. The goal is to reduce recovery time in similarly rare cases.

Despite the outage, Sui stressed that all core safety guarantees were preserved and that the network’s architecture prevented any risk to users or the blockchain’s state.