# round60 FIX delivery (testmgr2, 16 Oct, #2639)
Script: /opt/crew/work/testmgr2/r60_audit.py, table r60_audit.csv (per pool: exit, entry/exit time, LP removals after entry, rebuilt vs csv x).

## 1. Quiet exits vs pulls: CORRECTION of my 16 Oct note (#2635 said 3/37, wrong)
Joined the 37 quiet pools against v4tape/liq.jsonl AND /opt/ppcopy/lp_remove.db (to 16 Oct): **all 37 have a liquidity removal after entry**, almost always within seconds-minutes of the last swap (e.g. removal +218 s, last swap +228 s). So quiet = silent pull (rug), not "pool went quiet". Counting them 0x is CORRECT, not just conservative: x_if_quiet_sold 1.40x is NOT achievable. Fwd mean 1.474x (n63, 80% boot band 1.16-1.80) already has them at 0.
Residual risk is on the TP side (below), not on quiet.

## 2. Hand rebuild of fills (tape swaps, independent code, not round60.run)
All 11 tp trades rebuilt: entry = first swap >= first+63 s, exit = first swap with price >= 5x entry; every exit_t matches to the second, net = 5 x 0.9999 x 0.97 = 4.850 (csv 4.85). 
Anomaly: 5a9137ee458a: full position removal (-1.0e23, whole liquidity) at 1791810295.6, tp swap at ...304 (9 s after). Swaps keep printing with unchanged liq (tape liq field is the pool total at swap time, so another position kept trading or the removal tx is not the last LP). Worth a sell-sim look by auditor2. If that one trade were 0: fwd mean 1.397x. All other tp pools: removal comes 3386 s+ after the tp (after our exit).
Chain-level (tx receipt) rebuild not done; tape = chain events from our indexer.

## 3. Replay n>=50
Replay arm Q on tape to 8 Oct (busy60 pipeline, same code, pulls=0, quiet=0): n=261, mean 1.687x, 80% band 1.52-1.86; exits quiet 149 / tp 60 / time 51 / pull 1. Forward n63 1.474x (band 1.16-1.80): mean below replay's lower edge by 0.05, bands overlap. Caveat unchanged: the Q threshold (round-number share >= 0.5) was picked on that tape (in-sample). Forward is the out-of-sample check; forward n63 > 1.0 with lower bound 1.16.

## 4. Late flag
Works: trades.csv holds only on-time rows (late=0 by construction), 60 late rows are in trades_late.csv (live would not enter). Not a bug.

Remaining for VERIFIED: outside-seller / sell sim (auditor2) incl. the 5a91 case; nothing else open on my side.