The 14:45 UTC minute looked like ordinary order-book maintenance—327 cancellations, 296 limit-order placements, and 246 price updates—but one uncommon position-close transaction moved the story. Across all 1,649 transactions, 106 failed (6.43%), below the frozen corpus rate of 8.36%. The system structurally explained 1,474 transactions end to end; that does not mean they all succeeded.
The standout was 2SpV7Gc5, at checkpoint 300472588. Its asserted behavior was closing a liquidity position, a behavior appearing only 19 times among the 1,617 rows used for behavior counts. The transaction detail gives a concrete five-step sequence: construct an empty CETUS coin, collect a CETUS reward, construct an empty SUI coin, collect a SUI reward, then call the position-closing function. The sender’s boundary movements were +646,429.363121513 SUI, +11.448722702 CETUS, and +7,768.328342 USDC, with 0.0158428 SUI charged separately for gas. The detail does not assign the reward amounts to individual calls, and these movements establish neither profit nor an exchange price.
That one transaction represented about 65.7% of the minute’s gross SUI holding inflow. Across the whole minute, sender-boundary evidence recorded 983,829.28845509 SUI in and 10,063.351826506 SUI out in the holding record, for a net of 973,765.936628584 SUI. USDC showed 17,115.957528 in and 8,443.990417 out, while CETUS showed 119.995657435 in. Gas was separate: the minute’s gas record was a net outflow of 1.596297188 SUI. These records must not be added across evidence categories.
The failure population was fragmented rather than dominated by one outage. Fifteen failures stopped at a flash-swap function, fifteen had no recorded abort site, thirteen stopped at a pool swap, twelve at a price-freshness check, and eleven at a minimum-amount validation. One concrete contrast is 5MFvEhVo: it withdrew 506.771538 USDC, prepared the opposite SUI value, and reached pool::swap, where it aborted with Move code 1010. Six later steps never ran, and the application effects were rolled back; only 0.000217972 SUI of gas remained as a sender movement. That all-or-nothing behavior is also part of Sui’s transaction model. [Sui platform paper](https://docs.sui.io/paper/sui.pdf)
A committed price-update example, 4DyxZWXg, submitted observations of 0.753311600789661 USDC per SUI and 1.3277540980713259008 SUI per USDC with an explicit expiry. The calls ran, but the record carefully does not claim that an oracle value changed. This is the minute’s broader lesson: execution can establish a reached call, while the transaction detail and boundary movements determine what can responsibly be said about the outcome.
Three exact revisions also have their earliest stored calls in this minute, more than 100 minutes after observation began: 0x0b5f…c2edb, 0x82e6…aaf3f7, and 0x6665…219e19. Each appeared only once here, and their rows remained unrecognized. They enter this window's observed repertoire at 14:45, but that does not establish deployment time, and they do not explain the large position-close movement.
Useful exits are 2SpV7Gc5 for the position close, 5MFvEhVo for the aborted route, address 0xdc75a4…370c1a for the close-position sender, and pattern sui/close-liquidity-position/roster-f0137/v1 for the reusable transaction shape.