When a blockchain markets itself on speed, the harshest test is not how fast it runs — it's how it behaves when it breaks. Over 48 hours in late May 2026, Sui broke three times.
What Happened: Three Halts in 48 Hours
Between Thursday, May 28 and Friday, May 29, 2026, the Sui Layer-1 blockchain suffered three separate mainnet halts within roughly 48 hours, all ultimately traced to bugs exposed by the network's v1.72 upgrade path . The first two outages clustered around a new feature called Address Balances; the third had an entirely separate, latent root cause in distributed key generation (DKG) state persistence that only surfaced when validators restarted . As of June 9, 2026 Sui's live status page listed all systems operational but showed validators at 99.32% uptime over the trailing 90 days, a number that still carries the footprint of that downtime .
Quick Answer: Sui halted three times in ~48 hours on May 28–29, 2026, all tied to its v1.72 upgrade. No user funds were lost and no committed transactions were reverted, but SUI fell roughly 8% intra-cascade to near $0.90 and ended the week down about 19%.
The three failures had distinct signatures, and the middle one is the reason this incident drew unusual scrutiny:
- Outage 1 (~6.5 hours, May 28): The new Address Balances feature triggered an underflow crash. A transaction correctly cancelled for insufficient funds still entered "gas smashing," emitting a negative balance delta that was later applied against a zero balance — crashing validators .
- Outage 2 (~3.7 hours, May 29): Caused by the interim patch itself. The core team acknowledged it "accepted the risk" of a known low-probability halt path to restore service quickly — and that risk materialized .
- Outage 3 (~5.8 hours, May 29): A structurally separate, pre-existing DKG state-persistence bug surfaced at epoch change. Validators "forgot" that DKG had failed, and end-of-epoch logic stalled waiting on a completion that would never arrive .
On user safety, the Sui Foundation was firm: throughout all three outages, transactions timed out or were not accepted, but no user funds were at risk and no committed transactions were reverted on resumption . Markets were less forgiving. The SUI token fell roughly 8% during the cascade to a low near $0.90 and was down about 19% over the week .
The Deliberate Gamble: Knowingly Shipping a Risky Patch
The most consequential decision of the 48-hour cascade was a documented one: after the first outage, the Sui Core Team knowingly deployed an interim patch it knew carried a halt risk, prioritizing fast recovery over a fully safe fix. Per the postmortem, the team "accepted the risk accompanying this proposal in order to bring the halted network back as quickly as possible while a robust fix was developed" . That accepted risk became the second halt within 19 hours.
The interim fix pruned InsufficientFundsForWithdraw (IFFW) address-balance gas-payment entries before gas smashing, closing the underflow path that triggered Outage 1. But the team flagged an unresolved edge case in the same disclosure. As the postmortem explained:
"A transaction may have multiple reasons for cancellation, and one reason can override the others," — Sui Foundation postmortem (source: blog.sui.io).
In other words, a competing cancellation reason could mask the IFFW signal and bypass the protective guard. That documented path opened Outage 2 around 05:00 PDT on Friday, May 29, with more than two-thirds of stake upgraded to the next fix by 08:43 PDT .
The robust follow-up arrived as the emergency release mainnet-v1.72.4. Instead of pruning alone, an IFFW transaction now skips the executor pipeline entirely: it bumps mutable input versions, emits clean failure effects, charges zero gas, deletes no coins, and leaves the address balance at zero — no gas smashing at all . The team also added protocol-version discipline, gating the pruning behavior behind protocol v126+ and running a mainnet accumulator-version backfill at 692949576 so that transaction replay stayed bit-for-bit correct even though the hotfix landed before a protocol flag existed .
Outage 3 had a separate root cause and a separate remedy. When validators restarted at the next epoch boundary, distributed key generation (DKG) participation fell below threshold, but a pre-existing bug failed to persist the DKG-failure verdict to disk — so after further restarts, validators "forgot" DKG had failed, stalling the chain . The fix, shipped as mainnet-v1.72.5, added DKG-failure disk persistence plus a coordinated force_epoch_close operator lever that pins a consensus commit index to close a wedged epoch. Operators deployed it at epoch 1142, consensus commit index 768980, letting validators converge and resume block production .
Base Case: Sui's Reliability Record and Thesis Position as of June 2026
The base case for SUI in June 2026 is a high-throughput Layer-1 with intact core technology but a worsening reliability cadence. The triple v1.72 halt was Sui's third major mainnet incident since its May 2023 launch, and the gaps between failures are shrinking: roughly 14 months from the November 21, 2024 halt to the January 14, 2026 stall, then about five months to the May 28–29, 2026 cascade . An accelerating incident frequency is the single most important fact a position-sizer needs to weigh against Sui's technical roadmap.
| Date | Root cause category | Duration | Token impact | Fix release |
|---|---|---|---|---|
| Nov 21, 2024 | Congestion-control assert (TotalGasBudgetWithCap mode) | Single halt | — | Postmortem |
| Jan 14, 2026 | Validator divergence in consensus commit processing | Single stall | — | Postmortem |
| May 28–29, 2026 | Gas-charging + Address Balances bug, plus latent DKG persistence bug | 3 halts in ~48h | ~−8% intraday, ~−19% on week | v1.72.4 / v1.72.5 |
What did not break matters as much as what did. Sui's Move VM, object-centric parallel execution, and sub-second finality remained functionally intact throughout all three outages, and the foundation confirmed no user funds were at risk and no committed transactions were reverted on resumption . The very features that triggered the cascade — Address Balances and Gasless Stablecoin Transfers, both shipped in mainnet-v1.72.2 on May 20, 2026 — are meaningful UX advances once stable, removing the need to manage discrete gas coins .
The live status page told a quantified story: as of June 9, 2026 all systems were operational, but validators showed 99.32% uptime over the trailing 90 days . That figure sits above Solana's early-era reliability record yet materially below the five-nines (99.999%) availability institutional infrastructure buyers expect. For the base case, the recovery pace of DeFi TVL and transaction volume in the weeks after resolution is the live indicator to track — a fast rebound signals scar tissue, a slow one signals structural distrust.
One genuinely positive data point cuts the other way: validator coordination matured under fire. More than two-thirds of staked supply upgraded to each emergency release within hours — by 13:32 PDT for the first fix and 08:43 PDT for the second . A network that can push coordinated hotfixes across two-thirds of stake in single-digit hours has a functioning operational backbone — the base case is a maturing system with a code-quality problem, not a governance one.
Bull Case: Why Reliability Scars Can Heal
The bull case for SUI rests on a simple observation: each failure was followed by disclosure, not deflection. Sui published a detailed, mechanism-level postmortem naming the exact gas-charging edge condition, the distributed key generation (DKG) persistence gap, and even the pull request numbers behind each fix . That is a higher transparency standard than most Layer-1s offer at an equivalent age, and it gives traders an auditable record rather than a vague apology.
More importantly, the responses were structural, not cosmetic. Three architectural guardrails emerged from the cascade: the IFFW clean-exit path, which makes an underfunded transaction fail with zero gas charged instead of panicking during settlement, shipped in emergency release mainnet-v1.72.4 ; DKG-failure persistence plus a coordinated force_epoch_close operator lever, added in PR #26846 and shipped as mainnet-v1.72.5, let validators close a wedged epoch at a pinned consensus commit index and resume block production . These reduce the blast radius of future edge cases rather than papering over a single bug.
The forward-looking commitments matter too. Sui's own lessons cite cleaner and more exhaustively tested gas-charging logic, AI-assisted production debugging, and better failure containment so that a bad input can be skipped rather than halting the whole network . As the Foundation framed the priority in its postmortem, the goal is "better failure containment so a bad input can be skipped rather than halting the whole network" — Sui Foundation engineering team (source: Sui blog). That reads as sustained investment, not a one-time hotfix cycle.
Crucially, the damage had limits. No user funds were at risk and no committed transactions were reverted on resumption . The longest single outage ran roughly 6.5 to 7 hours — uncomfortable, but within tolerable downtime for gaming, NFT, and non-financial-settlement use cases that do not depend on second-by-second liveness.
Finally, the competitive moat. Aptos shares the same Move virtual machine but has seen slower ecosystem adoption. If Sui's reliability stabilizes, its developer and liquidity lead over the closest Move-native alternative is defensible — the scars heal faster when no peer is positioned to absorb the fleeing capital.
Bear Case: When the Network Is the Product
The bear case is structural, not emotional: for an infrastructure asset, reliability is the product, and Sui has now halted three times since its 2023 launch — a November 21, 2024 congestion-control assert, a January 14, 2026 consensus-divergence stall, and the May 28–29, 2026 triple halt . Three incidents in roughly thirty months is a compounding pattern, and each one resets the reliability-credibility clock with the enterprise and DeFi partners who underwrite Sui's long-term valuation.
The deliberate risky-patch decision is the sharpest data point for bears. By shipping an interim fix the core team knew carried a low-probability halt risk, Sui optimized for uptime minutes over protocol hygiene — and the second halt on Friday, May 29 was the direct, foreseeable consequence . The team's own framing makes the governance gap explicit:
"We accepted the risk accompanying this proposal in order to bring the halted network back as quickly as possible while a robust fix was developed," — Sui core team postmortem (source: Sui Foundation).
That trade-off is fatal to the institutional thesis. Enterprise protocols and institutional DeFi desks require service-level commitments that a 99.32% trailing-90-day validator uptime simply cannot support . Business-development conversations delayed or lost are a lagging indicator that will not surface in on-chain metrics for months. Meanwhile competing Layer-1s have a clean counter-narrative for sales cycles: Solana's materially improved uptime since its 2022 outages, and Aptos's halt-free launch record on the same Move virtual machine, can both credibly position against Sui's reliability story .
The price action validates the structural read. SUI fell roughly 8% during the cascade to a low near $0.90 and closed the week down about 19% . The bear thesis confirms itself if each successive incident produces a lower recovery high.
| Incident | Date | Root cause | Market reaction |
|---|---|---|---|
| Halt 1 | Nov 21, 2024 | Congestion-control assert (TotalGasBudgetWithCap) | Not disclosed in postmortem |
| Stall 2 | Jan 14, 2026 | Validator divergence in consensus commit processing | Not disclosed in postmortem |
| Triple halt | May 28–29, 2026 | Gas-charging + DKG persistence bugs (v1.72 path) | SUI −8% to ~$0.90; −19% on the week |
Portfolio Implication: Sizing SUI After a Triple Halt
The practical takeaway for a SUI position is to downsize, not zero out. The May 28–29 cascade was an operational failure, not an existential one: user transactions timed out but no funds were at risk and no committed transactions were reverted on resumption . A detailed, reproducible fix path — version-pinned releases through mainnet-v1.72.5 and PR #26846's DKG-failure persistence — separates this from an exploit or a consensus break with no recovery story . That distinction justifies a reliability discount, not abandonment.
The sizing framework is straightforward: hold any high-conviction SUI allocation at a discount until two consecutive major protocol upgrades execute without incident. The token's roughly 8% drop to a low near $0.90 and 19% weekly decline already priced in some of the damage , but the reliability question is unresolved while validators sit at 99.32% trailing-90-day uptime . The v1.73 deployment is the first observable test of whether the postmortem's lessons — exhaustively tested gas-charging logic and better failure containment — actually hold.
For traders leaning long, three catalysts argue for restoring full weight: a clean v1.73 rollout, TVL recovery to pre-incident levels, and major DeFi protocols publicly reaffirming their integrations after reading the postmortem. For the short side, three signals argue for cutting further: a fourth incident within six months, TVL migrating to competing Move-VM chains, or silence from enterprise integration partners that would suggest a stalled business-development pipeline.
The cleanest entry-timing signal is validator behavior on the next non-emergency release. Rapid, voluntary upgrade adoption — comparable to the two-thirds-of-stake speed Sui hit during the emergency patches — would indicate the operator set still trusts the code. Slow or coordinated-force upgrades would signal lingering doubt and warrant keeping the discount in place.
Concrete takeaway: treat SUI as a hold-at-reduced-weight through the v1.73 cycle. Re-rate upward only after two incident-free major upgrades and a TVL recovery; re-rate downward immediately on a fourth halt or visible partner attrition. The network's recovery was credible — the case for sizing back up is not yet proven .
Frequently asked questions
Were user funds at risk during Sui's three mainnet halts?
No. The Sui Foundation confirmed that during all three outages between May 28 and May 30, 2026, user transactions either timed out or were not accepted, but no user funds were lost and no previously committed transactions were reverted when the network resumed . The failures halted block production rather than corrupting state, so balances and finalized transfers remained intact once validators converged on a fix .
What is the Address Balances feature that triggered the first two halts?
Address Balances is a Sui feature, shipped in the mainnet-v1.72.2 release on May 20, 2026, that lets users hold funds as a canonical per-address balance alongside the older Coin<T> object model and pay gas with an empty payment vector instead of a concrete gas coin . The bug sat in the "hybrid" payment path: a transaction correctly cancelled for insufficient funds still proceeded into gas smashing, emitting a negative balance delta that crashed validators on underflow during settlement .
Why did Sui intentionally ship a patch it knew could cause another halt?
It was a speed-versus-stability trade-off. After the first outage, the Sui Core Team deployed a fast interim fix to restore service within hours rather than wait for the fully robust solution, explicitly stating it "accepted the risk accompanying this proposal in order to bring the halted network back as quickly as possible" . The interim patch carried a documented low-probability halt path that materialized as the second outage roughly 19 hours later, on the morning of May 29, 2026 .
How does Sui's outage record compare to other major Layer-1s?
Sui has recorded three full mainnet halts since its May 2023 launch, and the gaps are shortening — roughly 14 months to the November 21, 2024 halt, then about 5 months to the January 14, 2026 stall, then the May 2026 triple incident . That accelerating frequency echoes Solana's early-era reliability problems, while Aptos and Avalanche have fewer documented full halts. As of June 9, 2026, Sui's status page showed validators at 99.32% uptime over the trailing 90 days .
What should traders watch to assess whether Sui's reliability is improving?
Four signals matter most. First, whether the next major upgrade (v1.73) executes cleanly or triggers fresh incidents. Second, whether total value locked recovers to pre-May 28 levels after the SUI token fell roughly 8% during the cascade and about 19% over the week . Third, how quickly validators voluntarily upgrade on the next non-emergency release, versus the emergency patches v1.72.4 and v1.72.5 that resolved this incident . Fourth, whether major DeFi and enterprise partners publicly reaffirm their integrations after the postmortem.
