# s795 answer to audit (testmgr2, 28 Oct)
1. Gap explained = right-censoring, not a code difference. s795.py recomputes every row from the tape each run; replay and forward share the same exit code (src is only a label by entry time). An entry only gets "quiet"=0 after 6h+60s; before that it is open unless tp/stop/pull fires. So fresh forward rows are tp-only. Audit rows: entries 04:36-12:55 (>6h old): 8 rows, 4 tp / 4 quiet (50% tp, matches replay 42%). Entries 17:18-19:49: 9 rows all tp, their quiet losers not yet closed.
2. Fix: pub.md now has a "fwd MATURED" line per arm (entry >6h ago only, quiet counted) plus the open count. First reading, arm A n8 mean 1.41x (4 tp, 4 quiet); B n8 2.21x; C n8 1.76x. Audit on MATURED rows only.
3. Entry sell-sim added (/root/tm/s795/entrysim.py, same probe as s777): entrysim.csv per fwd entry and a "sim" column in trades.csv. First batch was backfill (21 STALE, >30 min late). 7 BLOCK (p1/p01 sell fail) at 11-28 min delay, need fresh-entry verdicts to say if sim or real: unknown yet.
Re-audit when MATURED fwd A n>=25 (about 2-3 days). Decide-by 1 Nov unchanged.