At 22:15 UTC, Sui processed 4,681 eligible transactions in roughly 60 seconds. The minute’s defining record was JCjNsTJb: a classified position-closing workflow whose sender boundary changed by +646,012.814069443 SUI, +295,112.077668 USDC, and +0.313295213 CETUS. That is boundary movement, not proof of profit; the transaction detail does not assign those amounts to individual reward calls.
The outlier sat inside an unusually concentrated, high-speed minute. Three senders accounted for 2,075 of 4,681 eligible transactions (44.3%), while the median gap was 0 milliseconds and the longest idle period was only 442 milliseconds. The three addresses ran repeated order, cancellation, pricing, and route-shaped activity at similar cadence; the evidence does not establish that they share an operator. A separate gas accounting view records 7.173872192 SUI in net gas for the minute, including 1.642471720 SUI from failed transactions.
The failure surge came at the end
The minute’s eligible failure rate was 935 of 4,681 (20.0%), compared with 8.36% across the frozen atlas. The change was chronological: 405 of the first 3,129 transactions failed, but 530 of the final 1,552 failed. In those last ten seconds, failures reached 34.1% of eligible traffic.
The largest named stopping point was flash_loan::return_flashloan_quote, with 103 of 935 failures (11.0%). Other common stopping points included a minimum-amount check (71), an internal flash swap (68), a pool swap (49), and a slippage verification (44). Among the 4,376 transactions usable for named-operation shares—after excluding 297 transactions that never reached execution and 8 contradicted records—764 failed. Of those, 233 were attempted swap routes, while 447 had no named whole operation.
That distinction matters. A route-shaped transaction is not automatically an exchange, and a reached call is not necessarily a completed effect. DfspDgKA opened a flash obligation, passed through alternating USDC/WAL and WAL/DEEP/USDC calls, then aborted at return_flashloan_quote. Its application effects were rolled back, leaving only 0.000462756 SUI in gas-related sender movement; Sui’s transaction model likewise treats Move aborts as all-or-nothing for application effects while still charging gas ([Sui paper](https://docs.sui.io/paper/sui.pdf)).
Orders filled the foreground, but not necessarily the book
The largest named activities were limit-order placement at 968 of 4,376 usable rows (22.1%), cancellation at 760 (17.4%), and swap-route attempts at 364 (8.3%). The order activity remained prominent even during the failure surge. For example, 6SdSUmib supplied a WBTC/USDC limit-order call with a stated price of 64,821 USDC per WBTC and an order size of 0.10879 WBTC. The record supports an order-placement attempt and a 0.00157674 SUI gas outflow; it does not prove that the order entered the book, was filled, or that the quoted amount was spent.
The minute therefore reads as two overlapping events: a large, concrete position-boundary movement early in the stream, followed by a late burst of flash-route and other transaction failures. The repetition is compatible with batched or automated execution, but these on-chain addresses do not identify a person, company, or bot.
For a closer look, start with JCjNsTJb for the position-close movement, DfspDgKA for the failed five-step flash route, and 6SdSUmib for the WBTC/USDC order-placement attempt.