THE DAY · 2026-07-19 · 22:12 UTC

One minute, one enormous position close, and 63 failed routes

2,380 transactions in this minute · 172 stopped early · browse every one →

At 22:12 UTC on 2026-07-19, Sui processed 2,380 eligible transactions. The standout was not a single successful exchange: it was a position-close transaction that recorded a sender-side increase of 646,012.523375258 SUI, plus 110,380.474608 USDC and 1.016186501 CETUS.

That transaction, H2MshUN4qYDTbThS7KBoujKiC9TgDZ3GrXFLrc3GrE7c, used two typed support values for reward-collection calls before invoking a pool position-closure terminal. Its own explanation says the exact reward amounts were not exposed, and the boundary movements cannot be assigned to a specific call. The evidence supports a close-position workflow and those sender-level movements; it does not establish profit or identify who ultimately benefited.

The minute was dominated by repeated order maintenance

After excluding 56 records from semantic analysis, 2,324 transactions remained for behavior counts. Of those, 598—or 25.73%—were classified as limit-order placement and 473—or 20.35%—as order cancellation. Another 220 updated prices. The three busiest senders contributed 368, 348, and 335 eligible transactions respectively, together accounting for 1,051 of 2,380 transactions, or 44.16%. Similar cadence and repeated structures show concentration, but they do not establish a shared operator.

The activity was nearly continuous: the median gap between transactions was 0 milliseconds, the 90th-percentile gap was 219 milliseconds, and the longest idle period was 449 milliseconds. Sponsored execution was also common, with another gas owner paying for 657 of 2,380 transactions.

The money tells a different story from the call counts

Sender holding movements in SUI recorded 2,751,470.292388483 SUI in and 449,229.548465546 SUI out, for a net of +2,302,240.743922937 SUI. The gas accounting was separate: 3.33238454 SUI out and 0.328658924 SUI in, for net gas of -3.003725616 SUI. These figures must not be added together because gas and holding evidence describe different accounting views.

USDC holdings recorded a net increase of 310,487.391602 USDC. DEEP recorded a net decrease of 67,571.920376 DEEP, while WAL recorded a net increase of 200,617.301980027 WAL. There are no prices in this frozen dataset, so these are boundary movements—not profit, loss, or cross-asset value.

The failures were concentrated in route attempts

There were 172 failed eligible transactions, or 7.23% of the minute. Among the 2,324 transactions usable for semantic analysis, 171 failed. That rate was below the frozen corpus-wide eligible failure rate of 8.36%, so this was not a minute-wide failure surge. The interesting failure pattern was narrower: 63 failed rows asserted execute-swap-route, representing 36.63% of all failures.

The most common stopping points were a pool swap with 22 failures, a minimum-amount check with 20, and flash-loan quote repayment with 15. Another 31 failures had no recorded stopping site.

8C1UM7wHEUf33C3fyk9ZoJnRadNXmjt236hZufqCDFjL shows the gateway case. It split an input, created a zero-valued companion coin, and reached gateway::swap_assets, but stopped at pool::swap with Move abort code 1010. The transaction recorded only -0.019872752 SUI at the sender boundary. The route was reached; an exchange was not established.

FyQikvebDvZBuT1pbcnuF5gPCw5vgxtBpg941R7KQAuS shows the flash route. It opened a quote-side flash obligation, passed through DeepBook and Bluefin, and attempted to return 63.732 USDC. It stopped at return_flashloan_quote with code 14680251. The application effects rolled back, leaving only -0.000367908 SUI at the boundary. Sui’s platform paper describes this all-or-nothing abort behavior, with gas still charged: [Sui platform paper](https://docs.sui.io/paper/sui.pdf).

The minute’s best-supported reading is therefore a heavily automated-looking order-maintenance stream surrounding a much larger position-close movement, alongside a set of route attempts that reached multiple venues but did not all complete. The records identify addresses and calls, not people, bots, intent, or realized trading outcomes.

Further trails: the position-close digest H2MshUN4qYDTbThS7KBoujKiC9TgDZ3GrXFLrc3GrE7c; sender 0xdc75a4952fc5b612ce967b14446d6067eafc768ae58819d371e9b83efd370c1a; and the reusable route pattern sui/execute-swap-route/coverage93-r004-committed-or-repayment-abort/v1.

Chapters are AI-authored exploratory readings of the stored minute — every figure in them is drawn from the same records the explorer serves.

← 21:55 the whole day ↑ 22:14 → this minute in the browser →