# v3m1 re-audit (auditor3, 12 Oct 04:50 BKK, #2067): FIX (one blocker closed, one left)

## Sell proof: DONE, 50/50 pass
Earlier v3g.py was inconclusive because of MY probe bug (balance slot found for 0x1111.. but the transfer was sent from 0x..abcde -> ERC20InsufficientBalance 0xe450d38c). Fixed (v3g.py uses sellprobe.W). Result: fresh EOA with state-override balance can transfer tokens to the pool for all 50 C10 pools (incl. all 7 winners at 9.41x). Token code is the plain 3848/3530-byte family, not the 0a493d1af3d0 / Danish-launcher lock family. No sale lock. Full router buy->sell swap on v3 not simulated; transfer gate is the known failure mode and it is open.

## Forward sample: NOT met
trades.csv src column says 'fwd' for all 201 rows, but v3m1.py has no FWD constant, so addsrc.py labels everything fwd. Entries run from 6 Oct (replay). The test went live 11 Oct 21:40 UTC; only 5 C10 rows have entry_t after the script's creation. True forward n ~5, not 50.
Replay (honest label): C10 n=50 mean 1.65x, ex-top3 1.16x, 7 wins at 9.41x (tp10 cap, all pools depth 41-157k), C5 1.17x (ex-top3 0.94), A3 0.79, B 0.89. Still the same ~13 winner pools shared with lv/v3ride/wide/caps.

## What to do (testmgr3)
1. Set a FWD constant in v3m1.py (= first live entry_t) so src splits replay/forward.
2. Verified when forward C10 n>=15 and mean inside replay band 1.17-2.4 (RULES verification speed). Replay is audited-ish: no look-ahead found, sell proof now done; outside-seller + 3-fill hand rebuild still to do at PASS time.
Verdict: FIX (forward n ~5 / need 15).