EscapistsReintroduced/docs/evidence/rescue-landing-2026-10-13

Faster rescue landing — 13 October 2026

The rescue pickup profile now commands up to 14 m/s horizontal alignment and 9 m/s descent while high. The descent command tapers continuously below 30 metres. The original final-ten-metre velocity, 15-metre horizontal alignment gate, two-second settled gate and three-second physically-ready hold remain in force. The default enemy/logistics helicopter profile, rescue escorts and outbound rescue landing profile retain their original velocity rule.

The isolated US run a02 completed 20 behavioral assertions and FIA run a03 completed 19, both with no script or VM failures. The two original Hueys landed on separate surveyed pads, accepted the original player and an already-unconscious living survivor into audited native cargo seats, then genuinely climbed above ten metres with those survivors retained, engines running, positive vertical velocity and no native ground contact. The single FIA Mi-8 separately passed the same landing, wounded boarding and loaded-climb checks; its native root compartment manager exposed all 14 free, audited cargo benches.

Transport	Height at landing entry	First physically ready, world seconds	Production boarding-ready, world seconds	Boarding-ready, wall seconds
US Huey 0	120.82 m	66.32	69.36	70.05
US Huey 1	150.10 m	74.42	77.45	78.14
FIA Mi-8	120.80 m	76.16	79.19	79.96

World durations start at the production m_LandingStarted timestamp. Wall durations start at the immediately following native observer log. Readiness sampling uses the production update cadence, approximately half a second. Landed() is observed only when production calls it; the fixture does not add calls that advance its readiness timer. Complete one-second flight samples and numerical readiness observations are retained in run-summary.json.

The comparable prior faction run a06 took approximately 152 wall seconds from its first sampled LANDING trace to the fleet physically-ready assertion. Its phase traces were ten seconds apart, so this is an approximate historical comparison, not an exact per-aircraft baseline or a controlled performance benchmark. Baseline observations retain the original log lines and source hashes. The new two-aircraft fleet was ready after approximately 78 wall seconds in this arranged local run.

At boarding readiness both Hueys had native ground contact, engines off, upward-axis Y above 0.9993, linear speed below 0.017 m/s and angular speed below 0.046 rad/s. This is stricter than the unchanged production limits of upward-axis Y at least 0.85, speed at most 0.5 m/s and angular speed at most 0.1 rad/s. The Mi-8 was also on native ground contact with engines off, upward-axis Y 0.9947, linear speed below 0.008 m/s and angular speed below 0.091 rad/s. The native contact path passed directly in all three aircraft; the sleeping-aircraft support fallback was not needed. Maximum sampled descent was approximately 8.584 m/s for the Hueys and 8.031 m/s for the Mi-8. The maximum lateral telemetry includes the initial phase-entry sample carrying the old cruise velocity before the first landing simulation update; it is not a claim that the new 14 m/s command caps inherited cruise motion instantaneously.

Fixture scope and diagnostics

The runner freezes production bytes into a new /tmp tree and launches a disposable local Workbench profile. Only the authored legacy test pickup is moved to previously verified flat airport grounds, the initial approach allocation is shortened to 400 metres, and the valid mission phases and signal are arranged. Native flight, collision, compartments, ground support and boarding/departure controls remain active; no aircraft is moved after allocation. Aircraft, original crew and the controlled player have combat-damage protection. The wounded survivor is a genuinely unconscious NPC registered under an isolated roster identity; loading uses the native cargo primitive after production physical eligibility and exact seat selection pass.

These are single-local-player integration tests. They do not qualify dedicated multiplayer, twelve simultaneous human players, remote casualty ownership or held-action RPCs, combat survivability, terrain outside the surveyed footprints, or the full escape-boundary journey. Landing time varies with entry height, simulation conditions and terrain. The tests do not measure server performance or promise a universal landing duration.

Passing behavior is distinct from a clean strict native report. US a02 retains 22 engine error occurrences and FIA a03 retains 23: each has 21 also seen in the previous a06 baseline, plus one (US) or two (FIA) suspicious-parent-velocity diagnostics for RPG trail particles during departure. No script/VM failure accompanied them, and both loaded-climb assertions passed afterward. The strict native report remains failed; complete console, script and error logs are retained. Run a01 caught an observer-only compile error because event is an Enforce keyword; the fixture parameter was renamed and the immutable failed attempt remains archived. Production code was unchanged between these attempts.

Final production hashes identify the two changed flight controllers and unchanged extraction/casualty dependencies. Comparing the complete pre-arrangement production manifests with the prior a06 baseline finds only those two controller changes. Every run includes its complete launch manifest, exact fixture, installed-addon input hashes and cleanup result. Both qualified runs confirmed production and installed-addon inputs unchanged and successful cleanup; a final read-only process check also confirmed no game process or active disposable prefix remained. Report binds the qualified results to these hashes. The runner only mutes its own local audio streams and never contacts or modifies the live game server.