At 13:03 UTC, Sui processed 1,956 eligible transactions. The minute’s defining event was not its failure rate—157 transactions failed, about 8.0%, slightly below the frozen corpus rate of 8.36%—but the contrast between relentless order-book maintenance and one enormous liquidity-position movement.
Among the 1,932 transactions with a usable behavioral explanation, limit-order placement led with 446 transactions and cancellation followed with 444. Together, those two activities accounted for 890 of 1,932 explained transactions, or 46.1%. One sender, 0xcde6db…6c0e04, submitted 484 of all eligible transactions, including 188 explained placements and 284 explained cancellations. The evidence shows a highly concentrated on-chain workload; it does not establish who controls the address.
The timing resembles scheduled or automated execution. The median gap between transactions was 0 milliseconds, the mean was 30.6 milliseconds, and the 90th-percentile gap was 209 milliseconds. Activity shifted in short bursts: during the first five seconds, placements led with 41 of 123 transactions; five seconds later, price updates led with 29 of 97; and a burst at 13:03:35 contained 368 transactions, including 86 cancellations and 64 placements. That cadence is consistent with automation, but the data does not identify a bot, operator, or shared organization.
The outlier inside the churn
Transaction 4jBFB2So arrived at 13:03:37.395 UTC from 0x672860…0318d3. Its explanation identifies a liquidity-position closure following an attempted reward-collection call. At the transaction boundary, the sender’s recorded net movement was +516,708.237748041 SUI and +125,230.087439 USDC. The separate gas accounting recorded −0.001524084 SUI.
Those are sender-boundary movements, not a profit calculation. The record does not attribute the positive balances to one specific call, and there are no prices for comparing the two assets. What it does establish is that this position-closing transaction was exceptionally large relative to the rest of the minute: its SUI movement was about 127 times the corpus’s per-asset 99th-percentile holding movement.
The store places 37 exact revision minima in this minute. Because the frozen corpus began at 13:01:48.968 UTC, only about 71 seconds of earlier observation precede them. This is still boundary accumulation: it identifies each revision's first stored call, not a change in the on-chain repertoire, a package deployment, a protocol launch, or a common operator.
What failed—and what failure means here
The largest named stopping point was flash_loan::return_flashloan_quote, where 38 failures stopped—24.2% of the minute’s failures. Another 36 failures had no named stopping point in the stored explanation, and 11 stopped at a minimum-amount validation check.
A representative example is DysAYeC6. Its two-call shape included owner-proof generation followed by a potential order cancellation, but execution stopped before any submitted command could be safely attributed because a withdrawal request exceeded available funds. The transaction therefore proves an attempted workflow, not an attempted cancellation that reached the order book. Since aborted Sui transactions roll back application effects, reaching a call in a failed transaction would not by itself prove that the call’s application change persisted.
The successful order-maintenance rows show the same need for precision. GQPQGcUg submitted 5,171 NS at a resting price of 0.01193 USDC per NS; the transaction’s detailed evidence does not independently establish that the order entered or remained on the book. In 5cafycKC, the explanation records a cancellation of 1,005 NS from the NS/USDC book at a 0.01194 USDC resting price. That is order maintenance, not proof that either order traded.
The minute’s story is therefore two-layered: nearly half of explained activity was concentrated in placing and removing orders, while a separate liquidity-position closure produced the striking SUI and USDC boundary movement. The repetition suggests a tightly scheduled workload; the evidence still leaves the controlling entities, economic purpose, and any realized trading outcome unresolved.
For a closer look, start with 4jBFB2So, the position-closure outlier; 0xcde6db…6c0e04, the minute’s most active sender; or sui/close-liquidity-position/d075-gateway/v1, the reusable explanation for the closure workflow.