Sui's mainnet halted three times in 48 hours on May 28-29, 2026, accumulating approximately 18 hours of total downtime. All three incidents traced to bugs introduced in the v1.72 software release, which added a new "address balances" feature designed to support zero-fee stablecoin transfers. The ...
"Outages suck... Iterating fast comes with pain." — Adeniyi Abiodun, Co-Founder, Mysten Labs / Sui
Sui's mainnet halted three times in 48 hours on May 28-29, 2026, accumulating approximately 18 hours of total downtime. All three incidents traced to bugs introduced in the v1.72 software release, which added a new "address balances" feature designed to support zero-fee stablecoin transfers. The cascade froze approximately $1 billion in on-chain assets, triggered $1.88 million in liquidations, and sent the SUI token down 19% on the week to $0.874 — 83% below its all-time high.
The incidents carry particular weight because they arrived less than four weeks after CME Group listed regulated SUI futures on May 4, 2026, and months after the SEC approved the first spot SUI ETFs. For a network actively courting institutional capital, three production failures in two days represents a material setback to the reliability narrative that underpins institutional allocation decisions.
No user funds were lost and no committed transactions were reverted. But the episode exposed systemic weaknesses in Sui's upgrade pipeline, failure containment, and epoch-transition logic — weaknesses the Sui Foundation has publicly acknowledged.
Outage 1 — Thursday, May 28, 13:48 UTC (~7:00 AM PT) Block production ceased. All validators entered a crash loop triggered by an underflow error in the gas charging logic. Duration: 5 hours 55 minutes. The network resumed at approximately 1:30 PM PT after validators adopted an interim patch.
Outage 2 — Friday, May 29, ~12:19 PM UTC (~5:00 AM PT) A masked variant of the same gas-charging bug bypassed the Thursday patch. The InsufficientFundsForWithdraw error code was overridden by a second cancellation reason, triggering the identical underflow crash. Duration: approximately 3.5 hours. A more comprehensive fix was deployed by ~8:30 AM PT.
Outage 3 — Friday, May 29, ~4:30 PM EDT (~1:30 PM PT) During the next scheduled epoch change, validators restarted to adopt the Friday morning fix. A latent bug in the distributed key generation (DKG) protocol's failure-state persistence surfaced: DKG failure status was never written to disk. Restarting validators were unaware that DKG had already failed, causing randomness-dependent transactions to queue indefinitely and blocking epoch closure. Duration: approximately 6 hours. Full restoration by ~7:20 PM PT.
Total downtime across the three incidents: approximately 15-18 hours.
All three outages originated from the v1.72 release, which introduced the "address balances" feature — a mechanism intended to simplify gas payments and enable fee-free stablecoin transfers.
Bug #1 — Gas Smashing Underflow: The v1.72 release introduced a hybrid gas payment system that combined traditional coin-object payments with the new address-balance model. When two transactions competed to spend from the same address balance simultaneously and one was canceled for insufficient funds (InsufficientFundsForWithdraw), the gas-smashing routine still attempted to spend those same funds. This produced negative balance deltas applied to zero balances, crashing the validator settlement process.
Bug #2 — Error Masking: Sui's transaction cancellation system allows multiple error codes to apply simultaneously. The Thursday interim fix only checked for the InsufficientFundsForWithdraw error code specifically. When a transaction triggered multiple cancellation reasons, a different error code took precedence, masking the insufficient-funds condition and bypassing the patch entirely. According to the Sui Foundation's post-mortem, the team was aware of this limitation when deploying the fix.
Bug #3 — DKG State Persistence: The on-chain randomness protocol uses distributed key generation across validators. A latent bug prevented the DKG failure state from being written to disk. When validators restarted (as required to adopt the Friday morning patch), each came back unaware that DKG had already failed during the current epoch. Transactions requiring randomness could neither execute nor be canceled, and the resulting backlog prevented the epoch from closing.
Each bug was technically distinct, but all three trace back to a single release cycle and the complexity introduced by the address-balances feature.
The Sui Foundation's post-mortem contains an unusual admission: the team knowingly deployed Thursday's interim fix despite understanding it carried a recognized risk of triggering another network halt.
According to the Foundation's report, the core team proposed a targeted fix at approximately 12:00 PM PT Thursday — skip gas smashing when a transaction is canceled with an InsufficientFundsForWithdraw error. Validators adopted it. The network came back by 1:30 PM PT.
The report states plainly that "the team accepted the risk accompanying this fix to bring the network back as quickly as possible." The predicted failure mode — error masking — materialized within 16 hours.
The decision reflects a tension common in high-throughput blockchain operations: the pressure to restore service rapidly versus the discipline to deploy only fully validated fixes. In this case, speed won. The consequence was a second halt.
Token Price:
Liquidations:
DeFi Ecosystem:
The outages arrived at an awkward juncture in Sui's institutional adoption timeline:
The timing gap between CME futures launch and the triple outage was 24 days. For institutional allocators performing due diligence on Sui as a CME-listed asset, the incident represents a data point that cannot be ignored. Spot ETF holders experienced a period where the underlying network could not process transactions — a condition that has no parallel in traditional equity or commodity markets.
A token unlock of 26.7 million SUI (0.27% of total supply, valued at approximately $27.9 million) to the Community Reserve was also scheduled for June 1, 2026, adding incremental sell pressure to an already-stressed token.
Sui's mainnet launched in May 2023. Its outage history:
| Date | Duration | Cause | |------|----------|-------| | November 2024 | ~2 hours | Transaction scheduling / congestion-control bug | | January 14, 2026 | ~6 hours | Consensus divergence | | May 28, 2026 | ~6 hours | Gas charging underflow (v1.72) | | May 29, 2026 (AM) | ~3.5 hours | Gas charging error masking (v1.72) | | May 29, 2026 (PM) | ~6 hours | DKG state persistence (v1.72) |
Four distinct incident windows in 2026 alone. Each with a different proximate cause, but two of the three May incidents sharing a common root in the v1.72 release cycle.
Cumulative estimated mainnet downtime in 2026: approximately 21.5 hours across four incidents.
Sui's outage pattern draws inevitable comparisons to Solana's early mainnet history. Between 2021 and 2022, Solana recorded 17 major outages, primarily caused by spam floods and excessive vote traffic. The comparison is instructive — but the trajectories are diverging.
Solana's current uptime stands at approximately 99.98%. The network has maintained 18 months of zero downtime as of May 2026. In December 2025, Solana survived a DDoS attack peaking near 6 Tbps without interruption. The Firedancer validator client — a ground-up C/C++ rewrite by Jump Crypto — went live on mainnet on December 12, 2025 and runs on over 20% of validators, providing client diversity that Sui currently lacks.
Sui operates with approximately 100 validators, compared to Solana's 1,500+. Single-client architecture means every validator runs the same codebase — a configuration where a single bug can and did crash the entire network simultaneously.
The question for Sui is whether its trajectory follows Solana's arc from frequent early failures to operational stability, or whether the ~100-validator, single-client architecture represents a structural constraint that limits reliability gains.
The triple outage exposes three structural concerns:
1. Upgrade Pipeline Risk: A single release (v1.72) introduced three exploitable bugs. The address-balances feature touched gas charging, transaction cancellation, and epoch-transition logic — three distinct subsystems. The blast radius of one release cycle affecting three subsystems suggests insufficient isolation between feature changes and critical protocol pathways.
2. Validator Homogeneity: With ~100 validators running a single client implementation, Sui has no client-diversity buffer. When the gas charging bug crashed one validator, it crashed all of them. Networks with multiple client implementations (Ethereum with Geth, Nethermind, Besu, Erigon, and others; Solana now with Firedancer alongside the Agave client) can contain the impact of single-client bugs to a subset of validators.
3. Epoch-Transition Fragility: The third outage — the DKG persistence bug — was latent. It existed prior to v1.72 but was only exposed when the post-fix validator restarts coincided with epoch change. This suggests the epoch-transition pathway has received insufficient adversarial testing. The Sui Foundation acknowledged this, committing to improved "end-of-epoch resilience."
The Foundation's post-mortem also flagged a need for better production diagnostic tooling, including AI-assisted log analysis, and enhanced failure containment strategies. These are process commitments rather than architectural changes.
Sui's triple outage is not existential, but it is material. No funds were lost. The bugs have been patched. The Foundation's post-mortem was detailed and transparent by industry standards. These are the marks of a team that takes reliability engineering seriously, even if the results do not yet reflect it.
The harder question is structural. A ~100-validator network running a single client implementation has an inherent fragility that software patches alone cannot resolve. Solana's reliability transformation required years, a second validator client, and a 15x larger validator set. Sui has not yet begun that process.
For institutional allocators evaluating SUI as a CME-listed, ETF-wrapped asset, the calculus is straightforward: the underlying settlement layer experienced 18+ hours of total downtime in a single week. Traditional market infrastructure operates under regulatory uptime requirements that make such incidents disqualifying. Blockchain networks are held to different standards — but the gap between those standards and institutional expectations is narrowing.
The economic value generated by a blockchain network depends fundamentally on its availability. When the network halts, all on-chain economic activity — trading, lending, liquidations, staking rewards — halts with it. The $1.88 million in liquidations and the $1 billion in frozen assets are the visible costs. The invisible cost is the institutional confidence that accumulates slowly and erodes quickly.