The first timing lead goes back to Polygon.

Start with the ten-minute lead

C91 deposited at 20:02:23 UTC on October 29. The next pool withdrawal was at 20:12:23, paying recipient 0xa7db…b6e3. Another deposit arrived at 20:11:35, just 48 seconds earlier. That deposit created the root used by the earliest withdrawals, so the root includes both deposits. Older deposits also remain possible; these are the two nearby anchors we checked.

Scroll across the chart or open it full size. Each box links to its transaction.

Open the full-size clickable chart ↗Timestamps and competing-deposit proof ↗Bridge and 100 POL transfer proof ↗

The ETH deposit and the 100 POL payment recover the same a7 signing key. Relay’s public record associates the exact source transaction and order ID with the Polygon fill; the fill receipt proves the actual POL credit. This links a7’s onward actions across chains. It does not establish that a7 spent C91’s note.

The 100 POL receiver is 0x828b51777329d3bab49880484f664f06df4fc9bb. Its current code is an EIP-7702 delegation indicator; no receiver authorization or controller identity was recovered here. Neither this address nor its implementation matched the selected Car or known incident addresses. At its own November 1 cutoff, a7 held 2,352.48530165 POL. That balance is not a complete accounting of this one fill.

How unusual is a ten-minute gap?

Among 143 other deposit anchors in the captured window, 142 had a later observed withdrawal; one remained right-censored. 70 of those 142 gaps were ten minutes or shorter. The median was 10 minutes 24 seconds. These are next-pool-withdrawal gaps, not known note matches. Several deposits can point to the same next withdrawal.

The ten-minute sequence is therefore a useful lead to follow, with little specificity on its own in this sample. We use later signed transfers and funding history to test it. We do not turn these counts into a probability that a recipient belongs to the attacker.

Download all 111 ranked withdrawal observations ↗

Five early recipients, five different financial histories

These are the five earliest distinct recipients. The rank is their withdrawal’s position in time. Rank 5 repeats the rank-4 recipient, so the fifth distinct recipient appears at rank 6. State observations below are pinned to November 1, 03:03:35 UTC.

Signed onward transactions and prior balances ↗ · History coverage and gaps ↗

There is also a fee pattern worth recording: ranks 2 and 3 are the only two of these 111 withdrawals whose declared fee exceeds half the 0.1 ETH denomination—93.29% and 99.27%. Most of their value went to the declared fee receiver, not the named recipient. The median fee was 2.11%. This is a descriptive pattern, without a tested ownership explanation. Both payment legs and fee distribution ↗

Test the Car timing against the available money

The first checked return to Car occurred at 21:48:24 UTC, 3 minutes 25 seconds after the fifth pool withdrawal. That proximity looks like a lead. The funding test supplies an alternative explanation: the returning wallet 0x7b259…4e4be already held 785,082.438836 pUSD before C91 deposited.

The retained canonical ledger has no incoming pUSD, USDC or USDC.e after that deposit through October 30 at 00:01:57 UTC. Its only three later transfers in those assets were the returns below.

785,082.438836 − 753,333 = 31,749.438836 pUSD. The returns were covered by existing cash; they did not require new principal from the later mixer withdrawals. Car’s earlier 750,000 pUSD transfers to this wallet occurred at 17:17, before the deposit and every candidate withdrawal.

This tests the proposed funding sequence. It does not resolve private intent, other assets, undiscovered accounts or ownership of a mixer note. Balance checkpoints and canonical transfer projection ↗ · Signed cash chronology ↗

What the timing pass establishes

It establishes an onward route for the earliest timing lead, exposes two large fee splits, and tests a specific Car funding hypothesis against available cash. It establishes no C91 withdrawal or Car financial connection. Earlier pool history and parts of the recipients’ indexed histories remain gaps after rate refusals. The large retained Car ledger was replayed locally; this download carries its hashes, checkpoints and selected authorization records, rather than all 262,361 source logs.

Research supports combining financial observations and testing heuristic assumptions; it does not provide a case-specific confidence score here. Tutela primary paper ↗ · Mixer-linking research ↗

Prepared 2026-11-01T03:23:10.052243+00:00. Records retain their individual block and query cutoffs.