# Sell probe for v2 0x8bce triggers (#2028, ppfeeds, 12 Oct)
Method: for the 31 pairs where the replay rule fired (Sync >=8x p0, reserve0 >=2 WETH; list trig.json), read all pair Swap logs from the trigger block to
min(end of tracked data, +60k blocks) and match each sell (amount1In>0) to the token Transfer(->pair) in the same tx. Tax per sell = 1 - swap amountIn / sum(Transfer value) (fee-on-transfer shows as less arriving). Data: probe.json, script probe.py (free RPC only, no spend).

## Result
- 25 of 31 triggered pairs: sells clearly work. 24-1208 outside sells each (total ~2,500 sells, ~2,500 ETH sold), tax 0.0% median AND max on 24 pairs. No token tax on any normal pair.
- 1 pair (0x16684607.., token 0xdc453b14..): median tax 0, max 99% on a single sell (one odd tx, likely a contract/blocked wallet), 639 other sells fine -> not a tax token.
- 6 of 31 are sell-dead after trigger: 0x8aa7e2c2 (12 buys, 0 sells), 0x03fef041 (4/0), 0xadf3539b (10/0), 0x5aeb6570 (30/0), 0x754b1044 (12/0), plus near-dead 0xf61498a3 (96 buys, 3 sells, 0.03 ETH), 0x1c6f36e9 (75/1), 0x674619bd (39/1). These are the honeypot / LP-pull "zeros" the replay already counts (3 zeros of 26), i.e. a ~20% loss-of-whole-position class that the replay prices in.
- Conclusion: no token-tax problem; sellability is fine for ~75-80% of triggers; the remaining ~20% cannot be sold (or LP gets pulled) and are already the zeros in the replay mean 3.59. Rule keeps the replay's zero-rate honest only if no extra filter is added.

## Caveats
Evidence is from on-chain swaps by other wallets, not a simulated sell from our wallet. A cheap live guard exists: at trigger time, require >=1 outside sell (amount1In>0) in the pair since birth; this would have skipped 5 of the 8 dead ones (they had 0 sells at trigger time too? not checked: need sells-before-trigger count, offered as follow-up).