EscapistsReintroduced/docs/current-status.md

# Current development status

Evidence checkpoint: 19 September 2026, through the archived 084026 companion restart run and separate 084040 vanilla dedicated control. Later running tests are not included. The primary deliverable remains an independently authored replacement for the abandoned Escapists base mission, integrating companion behavior through its interfaces. The first release targets vanilla Everon and up to twelve players, completing prison → scavenging/intel → radio defence → helicopter extraction without the original mod or dependent extensions.

The successor has five independent addon projects, its own scenario/resource identities, authored Everon sites and vanilla asset references. Ordinary Core and companion startup now work with real local admission. Scripted base-loop checkpoints and focused physical support tests exist, but this is not a released or fully accepted replacement. All seventeen release cases remain pending in [acceptance](acceptance.json); development checks record narrower evidence.

## Established development checkpoints

| Area | Recorded evidence | Limit |
| --- | --- | --- |
| Ordinary game admission | Regular, non-Workbench [Core](evidence/game-2026-09-19-082733/game-startup-report.json.txt) and [all-companion](evidence/game-2026-09-19-082938/game-startup-report.json.txt) runs load their headers/MissionSystems and admit the genuine local controller to a living canonical prisoner, with a fifteen-second ready interval and zero captured errors. | Diagnostic noSplash and standalone round-local roster key; no ordinary full route, normal splash, network identity or manual input qualification. [Admission design](local-singleplayer-admission.md). |
| Independent base loop | [Two Core rounds](evidence/native-2026-09-19-080656/report.json) pass 564 assertions, including radio feedback, extraction and restart, with zero errors. | That snapshot predates the standalone admission helper. The later [two companion rounds](evidence/native-2026-09-19-084026-restart-station-error/report.json) complete 96 assertions but retain two support-station catalog errors during restart; they are a failed runtime result. |
| Prison, equipment and medicine | [Prisoner fixture](evidence/native-2026-09-19-063949/report.json) places one controlled player and eleven physical characters inside the prison with exact kits and real clue transfers. [Medical fixture](evidence/native-2026-09-19-062406/report.json) demonstrates blood-loss treatment, consumed supplies and native consciousness recovery. | Physical actors are not twelve connected clients. Ordinary treatment input, animations, supply balance and navigation remain unqualified. The medical snapshot retains historical catalog errors. |
| Discovery and radio | [Partial document loss](evidence/native-2026-09-19-071654/report.json) verifies close physical discovery after partial knowledge and loss of all documents. Radio feedback/continuous defence and interruption have complete Core-loop checks. | Manual discovery, travel, inventory interaction, combat balance and remote feedback remain unqualified. No replacement clues are spawned after loss. |
| Player lifecycle and extraction | [Lifecycle](evidence/native-2026-09-19-081601/report.json), [recovery](evidence/native-2026-09-19-064406/report.json), [casualty loading](evidence/native-2026-09-19-075514/report.json) and [stock removal](evidence/native-2026-09-19-080058/report.json) have focused native checks. The actual local incapacitated body can be loaded into a reserved bench and removed alive beside the authored door. | Direct completion/helper admission is disclosed. Held input, remote owner handshake, competing rescuers, disconnect/death/deletion during loading and real reconnect remain unproved. |
| Physical infantry | [Patrol](evidence/native-2026-09-19-070919/report.json), [HQ preemption/resumption](evidence/native-2026-09-19-072838/report.json) and [pending-spawn cancellation](evidence/native-2026-09-19-073601/report.json) pass without creating response forces. | The delivered-order patrol test bypasses radio transit. Integrated perception/combat, interrupted reports and support losses need broader qualification. |
| Physical mortar | [First projectile](evidence/2026-09-19-mortar-first-shot/report.json) passes 48 assertions through the actual uplink/HQ/downlink/native activity path, consuming one original HE shell and stopping without duplicates. [Exact inventory transfer](evidence/native-2026-09-19-083810/report.json) passes 119 assertions and restores all 27 original shells to their original owners. | Observation is synthetic; impact is unmeasured. Transfer fixture bypasses radio/scheduling and observed a synchronous callback. Full salvo, post-expiry authorization, genuine late callback, perception and remote ownership are separate. |
| Dedicated compilation | [Regular Linux server](evidence/dedicated-2026-09-19-083213/review.json) compiles Core and reaches prison phase with zero players/36 AI inside a network namespace without external routes. | Four resource errors remain; [vanilla-only control](evidence/dedicated-2026-09-19-084040-vanilla/review.json) reproduces all four. No participant joined. This local-server route is diagnostic-only; public/authenticated hosting is not configured or qualified. |

## Immediate work

1. **Finish current-source regressions and investigate restart cleanup.** The companion loop reaches both victories but emits two `SCR_BaseResupplySupportStationComponent` catalog errors between rounds. Preserve that failure and obtain the existing bounded teardown diagnostics before deciding whether a scoped lifecycle correction is warranted. Current Core regression must include the new standalone helper; focused callback replay/fail-closed probing is now implemented and awaits native execution.
2. **Close physical support gaps.** A report ledger now prevents a cancelled/completed observation from repeatedly allocating mortar orders. A new expiry fixture and its native run must verify this while HQ remains active. Aircraft support needs a newly written flight driver: the extension inherited its flight base from original Escapists, so a namespace-only migration is insufficient. Installed native pilot input APIs have been identified; powered flight, actual fuel use and safe landing remain feasibility work. Logistics and rearming must retain finite physical stocks and losses.
3. **Qualify the ordinary playable route.** Check gate/equipment access, medical actions, finding the radio without fixture knowledge, defend/recover behavior, ground travel, real boarding/piloting/navigation and partial escape. Review authored layout and finite supplies from actual play. Preserve vanilla unconscious recovery and explicit administrator abandonment.
4. **Establish genuine multiplayer evidence.** The private dedicated runtime is installed and native empty-server startup is demonstrated. Real accepted client identities, joins/reconnects, remote ownership and replication remain outstanding; local bodies/callback injections are not substitutes.

## First-release qualification

Cover simultaneous/late joins in each phase, preload cancellation, reconnect and connection-ID reuse, reserved bodies, spent lives, death/unconsciousness/treatment, spectators and streaming cleanup. Verify remote task/map/radio/result updates, settings authorization, seat contention, partial escape, empty-server behavior, defeat and repeated restart with up to twelve connected players.

Collect matched original/successor workloads with repeated client-frame and server-tick measurements. No FPS improvement is demonstrated; the [comparison tool](performance-comparison.md) currently has synthetic regression evidence only. Complete provenance/dependency review, build/test the package under successor identities and satisfy the [release checker](release-qualification.md) with applicable source-bound evidence. No release package is published.

Original mission sources remain read-only. Hidden test displays, private profiles/prefixes and muted test audio keep engine work off the user's monitors. RHS, additional terrains, full original feature parity, alternate starts and distant-unit simulation remain later scope. Historical audits and evidence describe their own snapshots; the [validation ledger](validation.md) records both failures and subsequent fixes.