Sui Mainnet suffered three separate outages over two days, May 28-29, 2026, after bugs were introduced in the 1.72 software release. The network experienced downtime spanning roughly 12 hours in total, though developers stressed no user funds were at risk and no previously confirmed transactions were reverted.
First outage: a double-spend bug
The first outage began around 7 AM PT on Thursday and lasted until approximately 1:30 PM PT. It was caused by a bug where two transactions tried to spend from the same address balance simultaneously, leading to a negative balance and system crash. The Sui Core Team identified the issue and deployed a fix, bringing the network back online by early afternoon.
Second outage: error code edge case
The second outage hit Friday morning around 5 AM PT. An edge case allowed a different error code to overwrite the original cancellation reason and bypass the guard. The team deployed a more complete fix by 8:30 AM PT. But that patch didn't hold.
Third outage: epoch change stall
The third outage – the longest – stretched from approximately 1:30 PM to 7:20 PM PT on Friday. A latent bug surfaced during an epoch change: validators restarted but not enough participated in the shared random setup. The outcome was never saved to disk, causing the network to stall. It took nearly six hours to resolve.
No funds lost, but price dips
Throughout the outages, user funds remained safe and all confirmed transactions stayed final. Still, the market reacted. The SUI token price sat around $0.88 at the time of writing, down 2.57% in the past 24 hours. The token’s market cap stands at roughly $3.5 billion, ranking it 32nd among cryptocurrencies. The outages follow Sui’s earlier $10 million security commitment after the Cetus hack, raising questions about the network's reliability under stress.
Three areas for investment
The Sui Core Team outlined three areas they plan to invest in after the incidents: strengthening end-of-epoch resilience, refactoring gas charging logic to be more modular and testable, and building failure containment to prevent a single bad input from halting the entire network. The team provided no timeline for when these improvements would be implemented.
The question now is whether the fixes can keep up with the network's growth – or if another bug in the next release will trigger a repeat of this two-day nightmare.
