Include the current addon, terrain integrations, tools, regression tests and recorded development evidence.
3.2 KiB
Admission timeout and physical HQ evidence — 1 October 2026
A delayed client preload acknowledgement could keep an initial spawn request waiting even after its own server deadline expired. The finalization handler now sends an expired or invalid scene into the existing rejection path before waiting on native preload completion. It still requires the exact admission ticket and verified scene to assign an actor. Late completion cannot revive a rejected ticket. Actual scene-deadline and handler regressions reproduce the old wait at 60 seconds and verify rejection without assignment after the fix.
HQ now rejects a pending containment order when radio transit has made its motion evidence too old, weak or uncertain. Rejection preserves the previously delivered assignment. Planning uses the same valid-heading threshold as delivery, so a slow track below 0.5 m/s selects a search instead of repeatedly requesting unusable directed containment. The exact threshold boundaries and single delivery are covered by actual policy and planner-expression tests.
An eligible original convoy vehicle now reports its physical position instead of the possibly stationary native group entity origin. The position is copied at observation time and reaches HQ only through ordinary radio delivery. Lost, captured, replaced or invalid vehicle roots are never queried; their original-group status fallback and incomplete mounted-ammunition information remain unchanged.
Negative searches now require at least one original, attributed, fit soldier on foot within 40 metres of the search destination. A group entity at that location cannot substitute for soldiers physically searching there. Unavailable or unconscious groups reset the continuous 30-second quiet-search dwell. Original roster, perception, radio and delivered-report revision checks remain in place.
All 251 registered regression programs pass. The hidden, muted normal generated route completes 55 assertions and one round through lockpick, carried intel, radio defence interruption/restart, physical helicopter occupancy, victory and result display. There are zero script/runtime failures. Thirteen known installed-resource errors retain strict failure; this is scripted local integration, not human escape, remote-client or release qualification. Source/dependency fingerprints and private-process cleanup are confirmed. Review and retained regression evidence.
A reviewed mounted-transfer proposal is preserved as an unapplied patch. It handles only the second courier leg with an already-carried original magazine. Successful finite pickup, delivered HQ demand, worker/muzzle ownership, approach/reboarding and native swap verification must be integrated before it becomes gameplay. No unused transfer API was added to production and mounted rearming remains unfinished.
All 52 earlier unstaged working files are preserved apart from the explicit test registration; all 533 checked reference sources and existing packages remain unchanged. The current packaged playtest candidate predates these source fixes.