EscapistsReintroduced/docs/evidence/convoy-hq-coordination-2026-10-04

Convoy travel and attack coordination — 4 October 2026

This review changes convoy travel/attack orders and permits finite radio-delivered infantry flanks through contact. It preserves original physical units, capture/release ownership, ammunition gates, home guards and finite force budgets. See behaviour and limits.

Native fixture

Private Linux dedicated servers load Nogova and the same authorized external addons, inside separate network/PID namespaces with backend access disabled. No public server, AMP configuration or Workshop publication is changed. Each run uses a copied addon; source hashes accompany the logs. Runtime engine 1.8.0.13 compiles the addon scripts.

The fixture lets the production HQ start its finite supply convoy. It records six original vehicles and fourteen original passengers every five seconds. After a travel interval it spawns four protected US AI stand-ins and sends one labelled synthetic initial contact through the production convoy tactics entry point. Native waypoints, steering, dismounting and subsequent infantry movement execute in the engine. Both sides' test characters have damage handling disabled so early deaths do not obscure movement. Vehicles are not teleported by the fixture.

This tests convoy execution after contact; it does not qualify native detection, radio propagation or generic HQ infantry flanking end to end. The ordinary HQ planner/delivery/contact gates are covered by source-method regression tests. Random mission placement differs between native runs, so displacement is observational evidence, not a controlled speed/collision-rate benchmark.

Runs

• baseline: exact addon hashes from the previously qualified UAZ work, before this convoy change. Ninety seconds of travel, then three minutes after the test contact. Both original infantry teams dismounted, but their largest recorded on-foot displacements were only 13.4 and 12.4 metres.
• combat: new dismount/transport/priority orders, before the final travel corrections. Same fixture timing. All fourteen passengers dismounted; both teams moved over 350 metres while executing flank/search orders. No script errors were recorded. This establishes physical movement rather than merely successful waypoint allocation.
• arrival-stall: three minutes of travel followed by two minutes of combat. This iteration exposed a follower stopping between the 75-metre arrival boundary and the former 80-metre hold boundary. It is retained as a failed travel qualification despite the fixture reaching its timed end.
• final: repeats the longer fixture with the arrival gap removed, stable straight-road lookahead, fresh trails after reform and a safe fallback behind an already parked leader. The original six-vehicle column completes its first physical arrival and logs CONVOY DEPARTURE for leg 1. All fourteen passengers then dismount; the two team leaders move 166.1 and 159.0 metres on foot within the two-minute attack segment. No script errors or stranded-vehicle releases occur. The fixture completes and its private server is stopped. See metrics.json and final-events.log for measured outcomes.

Earlier diagnostic attempts under /tmp/erri-convoy-scene-native/ either failed the test-only callback signature, failed to advance the diagnostic HQ tick, or did not obtain native contact with their AI targets. They are not counted as successful attack tests. The final fixture explicitly labels its synthetic contact rather than treating absent perception as a tactics result.

The last source edit removes transport restoration from team release/teardown: only live regroup with a usable original truck may restore it. The final gameplay run predates that teardown-only edit; the exact final source is compiled separately, and the release, captured-transport and regroup guards are regression-tested. Its full addon manifest is source-sha256.json.

Automated checks and limits

arrival-red.log reproduces the arrival failure against the pre-fix controller. column-green.log exercises the fix, including 75.1/78/80/90-metre gaps with no recorded trail, straight-road target retention, final physical arrival, spacing, supply/intercept ownership, regroup trail cleanup and bounded stuck recovery.

The full repository suite passes (suite.log); the final addon also passes native script compilation with all 442 authored addon files matching the recorded manifest. Additional focused checks pass after the teardown correction. Together these cover transactional waypoint allocation, all-member dismount staging, vehicle-usage restoration boundaries, shared attack-space leases, fresh confirmed sighting gates, engaged-squad assignment/delivery, contact-resistant finite flank orders and existing hunter coordination.

The fixtures run headless: conclusions come from native entity positions and event logs, not visual inspection of rendered combat. They do not prove collision-free driving, cover-fire effectiveness, every modded vehicle/terrain, long-session pathfinding, or multiplayer replication. Native navigation can still refuse a route; existing retry/release policies retain physical survivors without teleporting or creating replacements. External addon localization/resource warnings remain present and are not claimed fixed.