At 19:29 UTC, a single on-chain address accounted for 500 of the minute’s 1,182 eligible transactions—42.3% of everything recorded. Its activity was tightly repetitive: 276 cancellation attempts and 224 limit-order attempts, all using two owner-proof DeepBook workflows. The same address also paid the gas for those 500 transactions. That cadence is consistent with high-throughput automated execution, although the address alone does not establish a bot, operator, or shared identity.
Of the 1,182 eligible transactions, 1,181 had enough evidence for a semantic label; behavior percentages below use those 1,181. The dominant labels were cancel-order at 335 transactions (28.4%) and place-limit-orders at 283 (24.0%). Together they made up 618 of 1,181 semantically labeled transactions, or 52.3%. The minute was busy almost continuously: the median gap was 0 ms, the mean was 50.7 ms, and the longest idle period was only 669 ms.
The repetitive core
The top address’s 500 transactions were all structurally explained and none carried a recorded execution error. Its 463 ordinary owner-proof calls and 37 repeat variants alternated between removing existing order records and submitting new orders. That looks more like rapid book maintenance than a simple burst of trading. The stored meaning for the placement workflow describes a placement attempt; it does not prove that an order entered the book or remained there.
One concrete example, BjveCnLY, passed an owner-proof value into a DeepBook cancellation call for the NS/USDC book. The transaction recorded a resting price of 0.01143 USDC per NS and resting size of 1,049.9 NS. Those figures describe the order being targeted, not an execution price or traded volume. The transaction therefore shows a precisely understood cancellation attempt, not proof that an exchange occurred.
The minute also contained 184 sponsored transactions, 15.6% of the eligible total. Sender and gas-owner addresses are separate facts in those rows; sponsorship does not identify who ultimately controlled or benefited from an address.
Where the uncertainty gathered
Only 53 of the 1,182 eligible transactions failed, a 4.5% failure rate—below the frozen corpus’s 8.36% eligible-transaction rate. But the failures were unusually concentrated in repayment and balance checks. Twenty-six of 53 failures stopped at one of three repayment boundaries: 12 at return_flashloan_quote, 8 at repay_loan_base, and 6 at repay_loan_quote. Another 11 stopped at the framework’s balance-splitting operation.
34otB5tp shows the distinction clearly. It assembled an 11-step route: opened a base-side flash loan, moved value through SUI/CETUS and USDC/CETUS helpers, then presented a resulting SUI balance for repayment. The route stopped at repay_loan_base with Move abort code 401. The transaction was understood well enough to describe the attempted route, but application effects were rolled back; its submitted 0.753453971 SUI is not evidence of a completed swap or profit.
A shorter route, 3XhWFNqs, opened a quote-backed flash loan, called Cetus and Bluefin swap helpers, and then aborted at return_flashloan_quote. Its attempted route touched MANIFEST and SUI, but the venue calls do not establish a committed exchange after the abort.
The exception with the biggest movement
The minute’s most conspicuous value outlier was not part of the 500-transaction order-maintenance stream. In 7Ev3baDo, boundary movements show an inflow of 991,594,211,789,926 raw MANIFEST units in the holding accounting and an outflow of 500.003065596 SUI in the holding accounting. Gas accounting separately records an outflow of 0.003065596 SUI. These accounting categories are separate and must not be added together.
That transaction ran two exchange-related calls and later returned a MANIFEST coin, but its broader purpose was withheld and the stored explanation explicitly leaves the swap outcome unresolved. The movements are real boundary evidence; they do not prove an exchange, price, profit, or beneficiary.
The clearest reading of this minute is therefore two-layered: one address supplied a near-continuous stream of owner-proof order maintenance, while a much smaller set of flash-loan and balance-management routes repeatedly reached fragile repayment boundaries. The chain shows the calls, their timing, and their proven movements. It does not identify the operator behind the address or turn attempted routes into completed trades.