83919ea0ae
The per-record instrument works and resolves individual squadrons. Over 240 s with the hunting pilot, seven loss events all landed on the same record, tracking one e007 Turret squadron from 18 craft down to 2 while deployed held at 41 and the global count fell 300 to 284. Losses come in steps of 2 after an opening drop of 4, which is unexplained and recorded rather than smoothed over. This weakens the frame-rate explanation from the previous iteration. At ~16.5 fps against a 30 Hz tick, game time runs at about 55 % of wall-clock, so route entries t = 90 and t = 120 land near 163 s and 218 s wall. The run reached 234 s wall, roughly 129 game-seconds, passing both, and no arrival occurred at either. "The runs were too short" no longer covers t = 90 and t = 120, though it still covers 170, 210 and 240 -- the run was cut at 240 s by the turn timeout rather than the planned 330 s, so t = 170 was never reached. It also sharpens the event-gated model into something testable. The squadron ended at 2, not 0, and no squadron has been eliminated in any run so far. If the trigger is a squadron being wiped out rather than merely damaged, every observation to date is explained: seven kills produced no arrival because they never finished anything off. Next is the cheapest decisive experiment yet available: run 60-90 s longer so that squadron reaches zero and watch for a 0 -> n in the following samples. The ~210 s title movie at boot remains the binding constraint, leaving about 350 s of observation per turn.