At 19:05 UTC, one transaction—7zL6M7tq—recorded a sender-side increase of 646,015.013619074 SUI, plus 244,765.27092 USDC and 4.019454149 CETUS. It is described as two reward-collection calls followed by a liquidity-position closure. The detail does not attribute those amounts to any individual call, so this is a boundary movement, not a profit calculation or a claim about which reward was paid when.
That single transaction explains most of the minute’s exceptional value flow. Across all 1,392 eligible transactions, sender-attributed holding evidence recorded 851,690.306303175 SUI in and 4,643.452226474 SUI out, for a net of 847,046.854076701 SUI. The gas evidence is separate: it recorded 1.617090978 SUI out and 0.14739168 SUI in. These figures must not be added across evidence categories. The large SUI movement in 7zL6M7tq was roughly 158 times the frozen corpus’s 99th-percentile per-transaction SUI holding movement.
The minute was mostly order maintenance
The transaction population contained 1,392 eligible rows. After excluding 38 rows where execution stopped too early or the classification was contradicted, 1,354 rows supported behavioral shares. The three leading behaviors were cancel-order attempts, 322 of 1,354; place-limit-order attempts, 294 of 1,354; and price-update attempts, 233 of 1,354. Together they accounted for 849 of 1,354 behavior-classified rows, or 62.7%.
One address, 0xcde6db…6c0e04, supplied 485 of the 1,392 transactions—34.8% of the minute—and also paid gas for all 485. Its activity was concentrated in 265 cancellation attempts and 201 placement attempts. That repeated shape is consistent with automated order-management activity, but the address alone does not establish a person, company, operator, or bot.
For a concrete example, 9bZka42M passed an owner-proof value into a DeepBook placement call at 19:05:59.987. The call supplied an order size of 1,043.5 NS and a resting price of 0.0115 USDC per NS. The resting price is an input to the order attempt, not an execution price, and the transaction detail explicitly does not establish that the order entered or remained on the book.
Failures were present, but not the minute’s main anomaly
There were 66 failed transactions among the 1,392 eligible rows, a 4.7% failure rate—below the frozen corpus rate of 8.36%. Of the 1,354 rows supporting semantic behavior claims, 47 failed, or 3.5%.
The failure mix was uneven. Twenty-two failures had no recorded abort site. Fifteen stopped at a Para repay_loan_base call, and 12 stopped at Pyth’s check_price_is_fresh. The top sender contributed 21 failures: 19 with no recorded abort site and two at DeepBook’s big_vector::leaf_remove. Five of its unrecorded failures appeared together at checkpoint 300544308, timestamped 19:05:59.290.
EWWYtDCU is one of that cluster. It attempted a two-step DeepBook cancellation shape, but execution stopped before a submitted command could be safely attributed because a withdrawal request exceeded available funds. No application action is established for that row, and its recorded gas was zero.
A different failure shows why reaching a route is not the same as completing an exchange. J6KpZ2sc opened a flash-loan-style route, moved through SUI, CETUS, and USDC helper calls, and then aborted at para::repay_loan_base. The application effects were rolled back. The transaction proves an attempted route and a failure at repayment; it does not prove a completed exchange or any economic benefit.
The minute’s real story is therefore a contrast: routine, highly concentrated order-management traffic surrounded one unusually large position-boundary movement. The ledger shows both the mechanical repetition and the exception, but it does not establish who controlled the addresses or whether any quoted order or route produced the intended economic outcome.
Useful exits are 7zL6M7tq for the position closure, 0xcde6db…6c0e04 for the dominant order-management address, and sui/place-limit-orders/deepbook-owner-proof/v1 for the repeated placement shape.