- Sui experienced three mainnet issues on May 28 and May 29 linked to its v1.72 release.
- Two of the suspensions stemmed from a gas billing error highlighted by the new address system.
- The third shutdown followed a hidden error that was triggered during a validator restart.
The Sui Foundation printed Its full review of the three mainnet outages that brought the network down on May 28 and May 29, 2026. The company said the outages were caused by two known bugs in the v1.72 software release.
The post-mortem confirmed that the first two suspensions shared the same cause, while the third suspension was triggered by a hidden error that appeared in the validator restart.
The first eruption began around 7 am PT on Thursday, May 28, and ended around 1:30 pm PT the same day. The second stop began at approximately 5 a.m. PT until 8:30 a.m. PT on Friday, May 29.
The third outage began around 1:30 pm PT on Friday and ended around 7:20 pm PT. The foundation said no users’ money was at risk in the three incidents and the network did not reimburse the incident when it restarted.
I’m notnetwork brand token, trading at $0.8776, down 2.6 percent per day, 15.7 percent over the past week, and 73.0 percent over the past year.
Gas Flare Bug Drives Sui’s First Two Outages
The v1.72 release added a feature called addresses, which gives users a new way to save money and pay for gas without using cash products.
Sui retail can now pay for gas using only addresses, cash products, or a mix of both, which the group calls hybrid gas.
For systems that pay with cash or natural gas, the rush hour breaks before gas is charged. This method combines all the input costs into a single cost that is deducted from the gas. This method handles both successful and completed tasks.
The cause of the first two outages was a gas leak. If the reservation attempted to withdraw addresses in excess of the budget limit, the attempt was blocked and the transaction terminated with an InsufficientFundsForWithdraw error.
The foundation said that the accident did not happen when the gas broke out on its own. Applying addresses to products produces deltas that are mediated by the system configuration process.
The damage was caused by a negative stream created by restricted-but-still-crushed gas that is being used at zero rate.
This situation can only happen when two businesses hit the scheduler at the same time and compete to spend money from an address that could not pay both.
The developer blocked one of them with InsufficientFundsForWithdraw to prevent withdrawals, but the canceled project still had withdrawals due to gas violations.
The Provisional Arrangement Came With a Known Threat That Caused a Second Stoppage
The arrangement that the main group wanted on Thursday afternoon stopped the machine from breaking gas when the operation was canceled by InsufficientFundsForWithdraw.
Enough developers adopted the patch to bring the network down at 1:30 p.m. PT, with the team acknowledging the known vulnerability associated with the process.
The Foundation said that the transition to natural gas is a difficult task. Address banks interact with money in complex ways. Updates must preserve all previous actions or use state gating, because nodes can fork while repeating previous actions under new assumptions.
Sui gas charging also includes security checks that prevent any service from creating or destroying SUI. Skipping the step that takes the money paid to the right place can cause accidents. Billing for expensive products also serves as a major part of the defense against denial of service.
The temporary fix had a bug that the team announced when it shipped. A sale can have several reasons for the ban, and one reason can override the others.
Transactions using bank addresses can be stopped because too many important items are waiting to affect the shared hot item, so it will not be stopped on InsufficientFundsForWithdraw when another transaction takes from the same address.
In this case, the InsufficientFundsForWithdraw error is masked by another error, bypassing the patch and causing the same flood.
A real-life incident that hit the network on Friday morning, leading to a second outage. The team was close to completing a permanent fix at the time and completed it in time to deliver the new patch to verifiers by approximately 8 a.m. PT. Enough applicants were picked up to bring Sui by 9:40 am PT.
Randomness State Bug Caused Third Sui to Stop
The network was up and running from 9:40 am PT until 1:30 pm PT on Friday, when the long time change did not end and the network went down for a third time.
The Foundation said the third suspension was due to a hidden error whose conditions were set by the original restart.
At the start of each session, Sui’s app developers run key generation, or DKG, a protocol that runs a random beacon that is used by events that rely on chain volatility.
DKG requires more participation than formal consensus. When participation is reduced, the default will limit itself for the duration it was designed for.
When the official resumed the installation on Friday morning, the participation of the DKG for the next period was not enough, and the protocol stopped itself. A latent error means that the error message is not written to disk.
When the restart was followed, every candidate came back unaware that the DKG had failed. Random dependent trades are expected to be executed or cancelled. It is acceptable not to maintain the record that DKG has failed, it will not happen.
The waiting line grew, and the last-minute ideas, which were supposed to drain the line before it closed, were left waiting at the DKG that would never arrive.
The preparation consisted of two parts. The first part fixed the bug and added a suggestion to continue having DKG on reboot.
The second phase added a mechanism that allowed verifiers to close a fixed period of time at a joint location. The team used this new method once to close the time frame involved. The network moved into the new time period normally, and stability was restored.
The Sui Group’s Take on the Three Endings
The foundation created four products for the week. The long-term resilience was the first, the group feels that the current safety net for climate change may be too narrow.
The foundation says the environment needs to improve waste management practices throughout the recovery process and improve the containment system to make it more efficient.
The second mode of transport involved the natural gas transport itself. The damage in parts one and two all came from bugs in gas billing, a fatal angle related to the address booking system, booking checks, and the scheduler.
The group said that the concepts are now so complex that side cases are difficult to solve by observation alone. Coming out of the event, the Foundation said gas charging should have the same code as Move VM or the Mysticeti consensus engine.
The third vessel was armed with AI. AI agents with access to the production environment, able to interrogate authorized documents, manage teams of teams, and collect metrics on demand, quickly identify issues within a week.
The fourth factor was lack of self-control. Damages in the first two parts were caused by some that the authorities could not properly repair. The foundation said the system does not have a deep security system that can withstand the explosion of such an accident.
