The Sui Foundation’s post-mortem on May 31, 2026 laid out the full chain of events behind the three mainnet halts on May 28 and 29. The most striking disclosure was not just the bugs themselves. The foundation said the team deployed an interim patch even though it already knew the fix carried a low-probability risk of causing another halt, and that exact failure appeared the following morning.
According to the report, the outages did not come from a single defect. Two separate bugs were responsible for the three disruptions. The first two halts were both tied to the v1.72 release, which introduced an “address balances” feature allowing users to pay gas directly from an address balance rather than a coin object. The feature was intended to simplify transactions and support gasless stablecoin transfers.
How v1.72 created the path behind the first two halts
The failure appeared in a concurrent transaction scenario. When two transactions tried to spend from the same address balance at the same time and one of them had to be canceled, the gas charging logic did not properly respect that cancellation. The network still attempted to charge gas for a transaction that had already been canceled. That produced a negative balance, or underflow, during validator reconciliation. Validators crashed, and block production stopped.
Mysten Labs said it removed that crash path with PR 26828, which prunes address-balance gas payment entries during early aborts caused by insufficient funds. But the foundation also said the first recovery used an interim patch rather than a fully tested long-term fix. The reason was speed. The team chose to restore the chain quickly while the durable fix was being built in parallel, despite a known issue in the temporary patch that had a low probability of triggering another halt.
That choice mattered. On Friday morning, the network hit a variant of the known issue and stopped again. The report said the second halt lasted about 3 hours and 30 minutes, shorter than the first because the recovery framework was already in place and validator coordination moved faster.
The third halt came from a different bug during validator restart
The third disruption had an unrelated root cause. When validators restarted to install Friday morning’s long-term fix, the restart process exposed a latent bug in how the network preserved randomness settings across restarts. Sui’s randomness protocol runs automatically at the start of each epoch, but at that moment the randomness initialization required a quorum threshold higher than consensus could provide.
A separate latent bug, present in the codebase before v1.72, then prevented the network from recovering randomness state after restart. The epoch transition could not complete, and user transactions froze once more. In other words, the third halt was not another version of the gas bug; it was an independent failure mode exposed by the unusual restart conditions created during emergency patching.
Mysten Labs addressed that problem with PR 26846, adding a deterministic force-close mechanism for stalled epochs. The recovery target was set to epoch 1142 and consensus commit 768980, allowing validators to converge on a consistent state and resume block production without reverting any previously committed transactions. The foundation said the force-close tool was used once during recovery and is now permanently available as an emergency mechanism.
What the report confirmed about safety and market reaction
The post-mortem ended with four specific assurances: the bugs have been fixed, user funds were not at risk, no settled transactions were reverted, and the force-close epoch mechanism now exists as a standing emergency option. Those points addressed the immediate operational questions raised by the outages.
Market data cited in the source showed SUI trading near $0.8789. During the halt sequence, the token fell roughly 8% to a low near $0.86. By Monday it was trading close to $0.90, though it remained down about 16% for the week.
The report also placed the incident in a broader reliability record. It was described as Sui’s third major reliability event since mainnet launched in 2023, after a two-hour transaction scheduling bug in November 2024 and a six-hour consensus divergence in January 2026. The root causes were different across all three incidents, pointing not to one recurring defect but to gaps in pre-deployment testing around edge cases introduced by new features and their interaction with existing consensus behavior.