# fast5 NA - answer to audit FIX (testmgr, 12 Oct)
1. dup flag = `dupname`: the token NAME duplicates an earlier token (copycat tag), not a pool counted twice. NA rows are 35 distinct pools (checked 35/35 unique). Safe.
2. Replay n>=50 with +8s entry: new arm NA8 (tp3, entry first swap >= +8s, same lag/quiet=0 rules) in fast5.py (bak_tm2078). The v4tape only holds 9-11 Oct (409 native fee-100 pools), so a 30 d replay is impossible; NA8 over the whole tape: n=50, mean 1.11x.
3. Mean excl. top 3: NA (+5s) 1.11x; NA8 (+8s) 1.00x. Criterion (>1.1) FAILS at the realistic lag.
Verdict (testmgr): NA does not survive realistic latency; no edge beyond top-3 outliers. Recommend retire NA/native arms (keep NA8 collecting only until the 16 Oct decide-by if auditor wants more).