At 19:20 UTC, Sui contained 1,286 transactions. One on-chain sender, 0xcde6db…6c0e04, supplied 506 of them—39.3% of the minute. The address is not an identified person or operator, but its concentration shaped the minute: cancel-order accounted for 322 of the 1,282 transactions used for behavior shares, place-limit-orders for 292, and update-price for 213. Together, those three behaviors made up 827 transactions, or 64.5% of that population.
The cadence was highly compressed. The median gap between transactions was 0 ms, the 90th-percentile gap was 225 ms, and the longest idle period was only 485 ms. That rhythm is consistent with batched or automated execution, but it does not establish who—or what—was operating the addresses. Of the 1,286 transactions, 242 were sponsored, meaning another address paid the gas.
The failures had a recognizable seam
Only 53 of the 1,286 transactions failed, a 4.1% rate—below the frozen corpus rate of 8.36%. But the failures clustered sharply. Thirteen stopped at para::repay_loan_base, another 13 at para::repay_loan_quote, and 12 at pyth::check_price_is_fresh: 38 of 53 failures, or 71.7%.
Transaction 2gGGtUri shows the shape of the problem. It opened short-lived flash credit, routed a SUI balance through calls touching CETUS and USDC, and attempted to return 0.753453971 SUI. The return call then aborted at repay_loan_quote. The stored explanation says application changes were rolled back; the only established boundary movement is −0.00034204 SUI, including gas. The route was attempted, but its exchange outcome remains unresolved.
The money tells a broader story
Holding movements recorded during the minute include a net +128.780605348 SUI across 463 transactions, −25.675719 USDC across 26, and −128.202134255 WAL across 16. Gas is reported separately: the gas evidence records a net −1.211568554 SUI across 1,044 transactions. These are separate accounting views and must not be added together; there are no prices or profit calculations here.
The largest single boundary movement was stranger than the dominant order activity. BVGct2Lk recorded a native transfer of 14,251,165,340,000 raw units of 0x76a49e…TAKE, alongside 0.000123104 SUI in gas. The record establishes an outward asset movement from the sender, not a payout, recipient identity, or economic gain.
The orderbook activity itself was concrete but narrower than the labels suggest. In AUtMomiD, an owner-proof flow reached a cancellation call for the NS/USDC book. The explanation records 15,789.5 NS at a resting price of 0.0114 USDC per NS as removed from the book. That supports a cancellation attempt and its recorded order parameters; it does not show that the order ever traded.
This minute therefore looks less like one unified market event than a dense coexistence of order maintenance, price updates, transfers, and repeated route attempts. The concentration is real; the shared operator, strategy, and economic outcome are not established by these records.
For a closer look, start with AUtMomiD and the sui/cancel-order/deepbook-owner-proof/v1 pattern, then compare the failed flash-credit sequence in 2gGGtUri with the large transfer recorded in BVGct2Lk.