b0fc1e6e7b
pilot.py with SYLPH_PREFER=e010, 300s, mission clock sampled throughout. RISES REPRODUCED: 004 -> 008 in (114.8s, 134.7s] and 008 -> 012 in (196.4s, 213.3s], both brackets containing the predicted 120 and 210. Second independent run, so the arrival half is now n=2. DECREMENT INCONCLUSIVE. The pilot fought properly -- 424 fire=1 samples, 865 with the target inside 1500 units, closest approach 79, target e010 throughout, hull untouched -- and 13 ADAN died (129 -> 116). The counter held 012 for 98s. That is NOT evidence against the decrement: the 13 dead were not identified by class, this pilot's kills historically skew to turrets, and the live e010 count ROSE over the run so attacker deaths cannot be inferred from it. Recording it as inconclusive rather than as a negative, which is the error this corpus has already logged twice. POSITIVE RESULT: at TIME 05:56.85, both arms sampled together, REMAINING OB = 012 while live UN_e010_ADAN_Attacker_S = 23. So the counter is NOT the live attacker head-count; it tracks a subset, and 12 is exactly the three A-route squadrons' membership (3 x n=4). Phase 1 fields only 16 e010, so 23 live means later-phase or F-route squadrons joined without touching the counter. Anomaly recorded, not explained: the enumeration reports 2 _Player entities at two distinct positions.