At 19:53 UTC, Sui processed 2,311 eligible transactions. 773 failed—33.45%, roughly four times the frozen corpus rate of 8.36%. The disturbance arrived immediately: 740 transactions landed in the first ten seconds, and 421 of them failed. The leading stopping points in that slice were flash_loan::return_flashloan_quote with 101 failures, transactions without a recorded application site with 96, and framework balance::split with 53.
This was not one uniform contract failure. Across the full minute, the largest failure groups were 222 without a specific recorded abort function, 181 at balance::split, and 146 at flash_loan::return_flashloan_quote. The evidence supports a burst of incompatible or incomplete route attempts, but not a single root cause for every failed row.
For semantic counts, 2,216 transactions remained after excluding 95 execution-only rows. Among those, the leading asserted operations were order cancellation—342 of 2,216—limit-order placement—327 of 2,216—and swap-route execution—278 of 2,216. The route label was unusually fragile: 204 of those 278 route-labelled transactions failed. Twelve cancellation transactions failed as well; no limit-order placement transaction in the minute carried an execution failure.
The first ten seconds also explain why the minute became hard to summarize. Of 421 failures there, 548 failures across the entire minute had no asserted operation label, so the evidence does not justify assigning them all to swaps or orders. The system structurally understood 1,611 of the 2,311 whole transactions, down from 94.9% in the previous minute to 69.7% here. That structural classification is not a success flag: a failed transaction can still be completely understood. Sui’s execution model rolls back application effects after an abort while still charging gas, so a reached call proves an attempt, not a persistent result. [Sui Smart Contracts Platform](https://docs.sui.io/paper/sui.pdf)
One enormous movement—and no claim of profit
The most striking value event was 4PVqDVR4, at 19:53:01.628. It was classified as a liquidity-position close after fee-and-reward preparation. The transaction recorded sender-boundary inflows of 330,047.406736896 SUI, 710,706.938541 USDC, and 0.546070044 CETUS. The SUI movement alone was about 81 times the corpus’s 99th-percentile per-asset movement threshold.
Its detail is more restrained than the headline numbers: the transaction prepared empty coins, ran claim calls, and then reached a position-closing call with zero minimum-output guards. The stored evidence does not reveal how much was collected by each call. These are boundary movements, not a demonstrated payout, profit, or economic return.
Across the minute, the application-holding record shows gross SUI inflow of 504,447.772828641 SUI and gross outflow of 13,029.94768437 SUI, for a recorded net of 491,417.825144271 SUI across 1,330 transactions. Gas is reported separately; those evidence categories are not additive, and no price or cross-asset valuation exists here.
Three transactions that make the sequence concrete
CpFo4dd1 prepared a coin split and a zero-valued companion before calling swapper::swap_assets; it aborted there with a Move error. Its only recorded sender movement was −0.000755952 SUI, also charged as gas. The detail explicitly leaves the exchange outcome unresolved.
HfvU8nzt shows a longer version of the same danger. It separated commission and residual amounts, routed the residual through a MAGMA swap, passed a USDC coin toward settlement, and then aborted at router::finalize. Its 148.860631 USDC minimum-output value was a guard, not proof of received output. The route attempt therefore does not establish an exchange.
By contrast, 4PVqDVR4 completed its position-close pattern and recorded the large boundary inflows above. That contrast is the minute’s central story: a high-value close succeeded amid a dense opening burst of route failures, while the ordinary DeepBook traffic continued around it. The busiest sender, 0xcde6dbe…6c0e04, supplied 476 of 2,311 transactions, but the failures were distributed across many senders; the address is an on-chain identity, not evidence of a person or bot.
For the next inspection, start with CpFo4dd1 and HfvU8nzt under sui/execute-swap-route/strict-remap-efdf8b06cd06/v1, then compare the successful close 4PVqDVR4 under sui/close-liquidity-position/roster-f0116/v1.