At 13:28 UTC, Sui processed 1,905 eligible transactions, but 339 failed—17.8%, more than twice the frozen corpus rate of 8.36%. The store places 11 exact revision minima in this minute, after more than 26 minutes of preceding observation. That is a change in the window's observed repertoire, not evidence that 11 packages were deployed then. It was not one sender's outage: the failures appeared across flash-credit, swap, vault, and ordinary transaction compositions.
The stream was nearly continuous. Transactions arrived with a median gap of 0 ms, a 90th-percentile gap of 213.7 ms, and a longest idle period of only 459 ms. The previous minute contained 1,476 eligible transactions; this one contained 1,905.
A crowded order-maintenance backdrop
For behavior counts, 1,884 transactions form the denominator after 21 rows with no execution evidence were set aside. The leading activities were placing limit orders, 327 of 1,884; cancelling orders, 326; updating prices, 220; and executing swap routes, 214. Order placement and cancellation together accounted for 653 of 1,884 behavior-bearing rows.
Two senders were nearly tied at the top: 0xcde6dbe…6c0e04 submitted 251 of 1,905 eligible transactions, while 0xffd4f0…d2f0ad submitted 250. That concentration is substantial, but it does not establish a shared operator. Gas sponsorship was also common: 463 of 1,905 transactions had a different gas owner from the sender.
The order activity was not automatically realized trading. The stored meanings describe limit-order attempts and cancellations; they do not turn every call into a fill or prove that an order remained on a book.
Where the failures stopped
The largest failure label was unrecorded, with 104 of 339 failures. That means the available explanation did not assign a more specific stopping point. Among named sites, 70 failures stopped at framework coin splitting and 52 stopped while returning a flash-loan quote. Those two sites alone account for 122 of 339 failures, or 36.0%.
Transaction 8tF1GjRC shows the shape of one failure. It contained 12 submitted steps beginning with a short-lived flash-credit opening. That first application call aborted with code 11 at pool::flash_swap_internal; the remaining 11 steps did not run. Its only sender-boundary movement was −0.000380356 SUI, including gas.
AXpT7GTx shows a longer route reaching much further. Its 18-step composition redeemed funds, passed through successive router and flash-swap calls, and reached a final minimum-output check of 3.770064 USDC. It then aborted at slippage::verify with code 42. The route was attempted, but the detail explicitly does not establish an exchange result; application effects were rolled back and only gas remained as a durable charge. [Sui’s documented transaction model](https://docs.sui.io/paper/sui.pdf) describes the same all-or-nothing behavior for aborted transactions.
The money moved, but not as a profit story
The sender-boundary accounting records large holding movements during this minute. In the SUI holding bucket, gross inflow was 1,385,472.532835435 SUI and gross outflow was 92,461.420275408 SUI, for a net of 1,293,011.112560027 SUI. USDC recorded a net inflow of 922,455.347902 USDC. Gas is accounted separately: SUI gas outflow was 3.165368378 SUI against 1.704599316 SUI of returned gas evidence. These buckets must not be added together.
The biggest individual movement was PoeGmu9Y, a committed close-liquidity-position transaction. Its sender-boundary movements were +642,337.156059757 SUI, +450,733.776016 USDC, and +8.614789503 CETUS, alongside −0.015842116 SUI in gas. The transaction alternated two reward-collection calls before closing a position, but the evidence does not assign the amounts to particular calls or establish profit.
There was also a committed route-shaped exception. 8dTqRFjV created a swap context, passed it through SuiDex and Bluefin limit-aware swap calls, and confirmed the context. Its boundary movements were −0.00024326 SUI and +0.001096 USDC, but the transaction detail says those movements do not independently prove an exchange.
The minute’s story is therefore a collision of high-throughput order maintenance with fragile multi-step routes: many attempts, several recurring guard failures, and large boundary movements whose custody and economic meaning remain narrower than their raw totals suggest. The clearest places to continue are AXpT7GTx for the slippage abort, 8tF1GjRC for the flash-credit failure, PoeGmu9Y for the large position closure, and the sui/execute-swap-route/coverage93-r147-whole/v1 pattern.