The striking event at 20:45 UTC was not volume alone. One transaction, GmkesN2j, recorded a sender-boundary increase of 646,017.136190402 SUI, alongside 327,367.262977 USDC and 5.764955278 CETUS. That SUI movement was nearly 158.5 times the frozen corpus’s 99th-percentile absolute SUI holding movement. It was a boundary movement, not evidence of profit or a priced exchange.
The transaction ran a five-step close-liquidity-position sequence: construct a temporary CETUS coin, collect a reward, construct a temporary SUI coin, collect another reward, then call pool_script::close_position. The reader detail says the collected reward amounts are not individually exposed, and the transaction-wide movements cannot be assigned to a particular call. The evidence therefore supports a staged position-close workflow and a large sender-side inflow, but not the claim that every displayed token came from the close itself.
That event sat inside a remarkably compressed burst. The minute contained 1,474 eligible transactions, spanning checkpoints 300570128–300570378, with a median gap of 0 ms and a mean gap of 40.6 ms. The busiest five-second bucket contained 221 transactions, and the stream occupied nearly the entire minute. The three leading asserted behaviors were order cancellation (256 of 1,380 behavior-labelled transactions), limit-order placement (239 of 1,380), and price updates (207 of 1,380). The top sender contributed 275 of 1,474 eligible transactions; that concentration identifies an on-chain address, not a proven person, service, or bot.
The money picture was similarly uneven. Across sender-holding movements, the minute recorded gross inflows of 1,478,842.188764036 SUI and gross outflows of 1,556.84950605 SUI. The largest individual movements were two liquidity-position closures: GmkesN2j brought in 646,017.136190402 SUI, while 5XEmsQzK brought in 570,018.665629071 SUI, 288,855.346613 USDC, and 4.909456308 CETUS. Both details describe reward preparation followed by a position-closing call; neither exposes the precise amount collected by each internal step.
Not every route with exchange-shaped calls was an exchange. BM6fgy2H ran a five-step path involving a Cetus adapter, coin merging, and two transfers. Its sender-boundary movements were −31,112,367.855326962 SharkCAT and −141.095500661 SUI, but the stored explanation explicitly leaves the swap outcome unresolved. The safe description is that it executed a swap-route workflow, not that it completed a trade or generated a return.
Failures were elevated but not dominant: 182 of 1,474 eligible transactions failed, or 12.35%, versus an 8.36% failure rate across the frozen corpus. Of those failures, 89 of 182 had no named abort site. Among the named stops, 20 reached validation::check_min_amount, 12 reached pyth::check_price_is_fresh, and 11 reached flash_loan::return_flashloan_quote. The minute ended with a particularly concrete failure: 6WRHe6fS attempted the same two-step owner-proof/cancellation shape seen elsewhere, but execution stopped before application action because a withdrawal request exceeded available funds. Its row establishes the failed attempt, not a cancellation outcome.
The best reading of this minute is therefore a high-throughput maintenance stream punctuated by large liquidity-position boundary movements, with a late burst of insufficient-funds failures. The data shows what was attempted and what crossed the sender boundary; it does not identify an operator, establish profit, or turn every route into an exchange.