Sui publishes post-mortem after mainnet stall halted transactions for six hours

Sui Foundation has published a detailed post-mortem explaining the cause of the mainnet stall that disrupted transaction processing on 14 January. It confirmed that the network halted as a safety measure to prevent inconsistent state finalization.

According to the Foundation, the disruption lasted for approximately six hours. An internal divergence in validator consensus processing caused it. 

During the incident, validators were unable to certify new checkpoints, leading to transaction submissions timing out while the network prioritized safety.

Sui consensus divergence triggered safety halt

The Foundation said the incident stemmed from an edge-case bug in the consensus commit logic that affected how conflicting transactions were handled under certain garbage-collection conditions. 

As a result, different validators derived different consensus outputs and attempted to execute incompatible candidate checkpoints.

When validators detected that more than one-third of stake was signing a different checkpoint digest, checkpoint certification became impossible. Validators then halted progress to avoid finalizing an inconsistent state.

“This is the intended failure mode for this class of issue,” the Foundation said, noting that the network is designed to stop safely rather than risk forks or irreversible inconsistencies.

No forks, rollbacks, or fund losses

Sui stressed that the stall was not caused by network congestion, transaction volume, or external threats. Throughout the incident:

• No certified state forks occurred
• No certified transactions were rolled back
• User funds were never at risk
• Network safety and consistency guarantees were preserved

While transaction execution halted during the incident window, read operations continued to serve the last certified state. This ensured data consistency for users and applications.

Improvements planned after incident

The Sui Foundation said it is implementing several changes to reduce recovery time in the event of similar issues in the future. 

Planned improvements include faster detection of checkpoint inconsistencies and more automated operator tooling to clean up divergent internal state. Also, expanded consensus-specific testing to reproduce and validate fixes before deployment.

The Foundation added that while the interruption was disruptive, it confirmed that Sui’s safety-focused architecture behaved as designed.

Final Thoughts

• Sui’s explanation confirms the mainnet stall was the result of a consensus edge case, with safety mechanisms halting the network to avoid inconsistent finalized state.
• While disruptive, the incident highlights the trade-off between availability and safety as high-throughput networks push performance limits.