At 13:06 UTC, Sui handled 2,237 eligible transactions with a median gap of 0 ms. The minute recorded 143 failures, or 6.4%—below the frozen corpus rate of 8.4%. That makes this look less like a chain-wide outage than a concentrated failure family.
One sender, 0x7744ad…a7c7fa, supplied 90 transactions, including 65 failures. Those 65 failures were all asserted as execute-swap-route, accounting for 45.5% of every failure in the minute. They did not all stop at the same check:
- 32 stopped at
pool::flash_swap_internal. - 18 stopped at
validation::check_min_amount. - 10 stopped at
pool::swap_in_pool. - 5 stopped at
pool::swap.
The address also submitted 16 deposit-assets transactions and nine rows where no operation was asserted, so this was not simply 90 identical retries. The record identifies an on-chain sender, not a proven bot, operator, or team.
The route failures had a recognizable shape
DG5zu5a3 shows the minimum-output branch clearly. It withdrew 10,112.507021701 SUI, prepared a balance-manager proof, and attempted a 10,109.9 SUI limit order priced at 0.74922 USDC per SUI. It then reached a verification step requiring a minimum output of 1,514,988,660 raw units and aborted there. The submitted minimum is a guard, not evidence of output received. The remaining three steps did not run.
7HWGbHT9 failed earlier. Its first call attempted a short-lived flash-credit opening and aborted inside pool::flash_swap_internal; the remaining 11 submitted steps never ran. Both records show only the gas charge at the sender boundary. Sui transactions have all-or-nothing execution semantics: application effects from an aborted transaction do not persist, although gas is still charged. [Sui’s system paper](https://docs.sui.io/paper/sui.pdf) describes the same execution model.
The two transactions therefore prove attempted routes and their stopping points—not completed swaps, repayment, or economic outcomes.
Most of the minute was order-book maintenance
Across the 2,198 transactions where the system asserted an operation, cancel-order led with 526 rows, or 23.9%, followed by place-limit-orders with 507, or 23.1%. Together they represented 1,033 of 2,198 semantically classified transactions. update-price added 212 rows and execute-swap-route 163.
The largest sender overall, 0xcde6db…6c0e04, supplied 553 of the 2,237 eligible transactions, or 24.7%. Its representative transaction, GtRZNHDa, ran an owner-proof cancellation in the WAL/USDC book. The detail binds a 0.029422 USDC-per-WAL resting price and 170 WAL of resting size removed from the book. It does not establish that the order traded.
Fourteen exact revisions have their earliest stored calls in this minute. With observation beginning at 13:01:48.968 UTC, that means only that the corpus had not recorded them during its first four minutes; it does not establish a deployment or broader on-chain repertoire change. The timing also does not show that those revisions caused the route failures.
A large movement, without a profit story
The minute’s standout movement was CEn7hiLz, which ran reward collection followed by a liquidity-position closure. At the sender boundary, the stored record shows +515,208.944187532 SUI and +89,384.0554 USDC. The detail does not allocate those movements to one individual call or prove a reward, payout, profit, or beneficial ownership.
Across all holding evidence in the minute, SUI showed gross inflows of 2,029,672.640345745 SUI and gross outflows of 2,147,947.012998611 SUI, for a per-asset net of −118,274.372652866 SUI. USDC showed a net of +842.21263 USDC, while WAL showed +363,043.181982736 WAL. These are separate asset totals; they are not prices, cross-asset value, or profit.
The clearest reading is a high-throughput minute dominated by established order maintenance, with one sender repeatedly attempting several kinds of route and validation paths. The evidence shows where those attempts stopped. It does not establish why the sender kept retrying, who controlled it, or whether any intended exchange succeeded.
For a closer look, start with 7HWGbHT9 and DG5zu5a3 for the two failure paths, CEn7hiLz for the large boundary movement, or the reusable pattern sui/execute-swap-route/wave4-r1351-opener-abort/v1.