For roughly ten and a half hours on 2026-07-19, this address did one dominant thing: it repeatedly launched routed swap compositions, often only milliseconds apart. Across all 5,934 recorded transactions in the frozen window, 5,391 failed. The median gap was 209 milliseconds, the longest failure streak was 234 transactions, and the longest successful streak was only five. That rhythm is consistent with automated retry or route generation, but the evidence does not establish who controlled the address or whether any route was profitable.
The address paid its own gas in every recorded transaction. No transaction was sponsored, and the gas owner was the sender itself.
Two different views of the activity
The whole-history record contains 5,934 transactions. MotiveLayer could explain the complete structure of 3,514, explain connected portions of 2,368, and partly explain 52. This structural reading is separate from execution outcome: a transaction can be completely understood and still fail.
The detailed address stream is a separate, capped latest-500 sample, covering 2026-07-19 22:27:57–23:37:35 UTC. In that sample, 393 transactions had complete structural explanations, 105 described only connected portions, and two were partial. Exactly 433 of the 500 failed. The latest sample therefore shows the same character as the full history rather than representing the entire history itself.
What it kept attempting
Among the 5,752 whole-history transactions with enough semantic evidence for behavior statistics, 3,512 asserted execute-swap-route. Two more asserted native-coin consolidation. The remaining rows did not receive a broader behavior claim.
The route patterns repeatedly describe flash-credit or flash-loan openings, balance extraction, conversions between coins and balances, venue-specific swap helpers, repayment, and residual consolidation. That is a useful description of control flow, not proof of a settled exchange. In particular, the authored meaning for roster-f0098/v1 describes a wrapper borrow, a market-order adapter, a flash close, a reverse conversion, and consolidation—but explicitly does not establish a filled order or resulting amount.
The latest-500 stream makes the repetition visible: 393 rows asserted execute-swap-route, while 107 withheld a broader behavior claim. The stream contained many distinct route variants, but they were variations on the same flash-backed, multi-call theme rather than evidence of a different operating identity.
Where the attempts stopped
Across the whole history, failures clustered tightly around repayment and balance handling:
- 2,165 stopped at the framework balance-splitting function, 40.2% of failures.
- 1,447 stopped at quote-side loan repayment, 26.8%.
- 1,311 stopped at base-side loan repayment, 24.3%.
- 100 stopped while validating order inputs, 1.9%.
Those first three locations account for 4,923 of 5,391 failures. The latest-500 stream repeats the pattern: 215 balance-split failures, 102 quote-repayment failures, 82 base-repayment failures, and 21 order-input validation failures.
A representative whole-history transaction, CwVHYij, reached eight connected calls and submitted 5.766503889 SUI to a route before aborting at repay_loan_base. The detail says the application effects rolled back; the submitted amount is not proof that the value was spent or that an output was produced. Another, 9ZcEVAyX, reached a wrapper borrow, a market-order adapter, a flash close, and reverse conversion before stopping at balance::split. It paid 0.001229064 SUI in gas, but the attempted route does not establish an exchange.
The address was not incapable of committing. BwH6pXfN committed five connected calls and produced a proven sender-side increase of 0.044639401 SUI, but its own explanation still says the execution does not establish an exchange. A committed route is therefore not automatically a filled trade.
The money, without pretending it is profit
Whole-history sender movements show two separate SUI views. Gas accounting records 9.421726002 SUI out and 0.03891916 SUI in, for net gas of -9.382806842 SUI. Sender-holding accounting records 20.963872882 SUI in and 8.736154701 SUI out, for net +12.227718181 SUI. These figures are reported separately because gas can also appear as a holding change; they must not be added together.
Other whole-history holding movements were +2.507864 USDC, +9.643175 DEEP, +0.258861241 WAL, +394079121 raw units of ALKIMI, and +0.009988 USDSUI. The latest-500 stream shows +0.215368 USDC, +0.009988 USDSUI, and +23,405,795 raw WAL units alongside its SUI movements. There are no prices or cross-asset valuation basis here, so these boundary movements do not establish profit, loss, or a completed exchange.
The address’s busiest minute was 18:08 UTC, with 269 transactions. Its longest idle period was about 11 minutes 30 seconds, an unusually large pause compared with the sub-second bursts surrounding it. The whole-history activity began at 13:02:33 UTC and ended at 23:37:35 UTC.
The cleanest exits are 9ZcEVAyX for a late balance-split failure, CwVHYij for the repayment failure that rolled the route back, BwH6pXfN for a committed but economically unresolved route, and the reusable pattern sui/execute-swap-route/roster-f0098/v1 for the mechanism behind the repetition.