At 17:01 UTC on July 19, 2026, Sui produced a near-continuous burst of 7,022 eligible transactions. The median and 90th-percentile gaps were both 0 milliseconds, with a longest idle period of only 416 milliseconds. The pattern is consistent with batch-like automated execution, although the evidence does not identify who or what operated it.
The minute’s failure rate was 2,170 of 7,022 transactions, or 30.9%, versus 8.36% across the frozen corpus. For behavior labels, the denominator is 6,652 transactions with enough execution evidence for a semantic explanation. Their leading activities were execute-swap-route, 1,034 of 6,652; place-limit-orders, 1,027 of 6,652; and cancel-order, 1,009 of 6,652.
The failure burst had a center
The address 0x7744ad…a7c7fa submitted 774 of the minute’s 7,022 transactions and failed in 600 of them. Those failures were not all the same error: 206 stopped at a minimum-amount check, 182 at a flash-swap opener, 86 at a pool swap-internal call, 61 at a pool swap call, 34 at a square-root-price guard, and 31 while validating order inputs. Together, those six sites account for all 600 failures from that sender.
A representative attempt, 4ZVYVLTW, opened a partner-attributed flash route, read a repayment requirement, withdrew a repayment asset, prepared SUI and USDC balances, and reached a minimum-output check. It guarded on 626743465; that number is a condition, not proof of received output. The check aborted, application effects rolled back, and two later steps did not run. The transaction’s recorded SUI boundary movement was −0.000276498 SUI, also recorded as gas; those accounting views should not be added together.
Another route, DbrjDyk3, reached a directional CETUS-to-USDC router leg and returned the flash receipt to its repayment call before aborting at balance::split. Its detail explicitly says that reaching the router does not establish a completed exchange. The route was attempted; its application effects did not persist.
The concentration was substantial but not exclusive. The same minute also recorded 236 failures from 0xcde6dbe…6c0e04, 175 from 0xf95d2f…bcd057, and 115 from 0x9cad98…b7f6c2. Four addresses therefore accounted for 1,126 of 2,170 failures, or 51.9%, without establishing that they share an operator.
The exception was a large position closure
This was not simply a minute of broken swaps. At 17:01:50.526, Cx1uJpEb successfully ran a position-closing workflow with two reward-collection calls followed by the closing call. The transaction recorded boundary movements of +562,303.522979014 SUI, +32,540.848537 USDC, and +0.395135382 CETUS for its sender. The stored detail does not expose the individual reward amounts, so those movements cannot be converted into a profit claim or assigned to a particular call.
That one SUI movement was roughly 138 times the corpus’s per-transaction 99th-percentile holding movement. At the minute level, recorded SUI application-holding movements totaled 936,796.657819285 SUI in and 723,871.571627769 SUI out across 3,606 transactions, for a net of +212,925.086191516 SUI. Gas accounting separately recorded −12.922613502 SUI across 5,118 transactions. These are separate evidence categories, not a price-based return calculation.
The minute’s story is therefore two overlapping systems: a dense, highly repetitive execution stream in which one sender generated a large, varied failure cluster, and a smaller set of successful position, order, and asset-management operations that still moved substantial on-chain balances. The failures show exactly where attempted routes stopped; they do not show that the attempted swaps or repayments completed.
Trace the minimum-output failure at 4ZVYVLTW, the position closure at Cx1uJpEb, or the late flash-route failure at DbrjDyk3. The recurring route pattern is sui/execute-swap-route/wave4-r1351-minimum-guard-abort/v1, and the concentrated sender is 0x7744ad…a7c7fa.