Return Fire native objective: unmet; lifecycle qualification failed

Run: scratch/runs/20261109-return-fire-native-objective-1727/.

No enemy objective or new objective consequence was established. I personally reviewed the current raw bunker map and helicopter reconnaissance screenshots. They show the home bunker, a green-roof roadside building, coastline, palm-only islets and open water. The building's enemy identity was not established. No pickup, target destruction or flag consequence is claimed. This is a failed objective search, not another accepted movement outcome or evidence that the demo lacks enemies. No demonstrated emulator defect justified a code change.

Retained prerequisites and actual identity

All 700 paths in the accepted November 8 ledger existed locally. All seven retained qualified-control references and three inherited helper references matched their hashes. The accepted selector, actual consumer observer, bound input and numbered ACK path were reused; cancellation research was not repeated. Current source/build differed from the historical ledger in 34 paths. Before allocation, the new run recorded current hashes and executed the actual driver with DRIVER_PREFLIGHT=1: 698 pins passed, without Chrome or an original launch. The spent historical run and its pins were not overwritten. The owner disk floor of 1.5 GB replaced inherited 5 GB checks; input, deadline and identity guards were retained.

Tested source is 49066461ef486cb8393c3b896354f7d2006a131e plus the retained relevant-source.patch; Git and small helper patches are the source archive. The actual recovered guest module is 3e491c0840bb7771328738a0d160b4bbaa3d7ab72074cbd38a7b41e5d7e8ec18, matching the staged module pin. All 58 original consumers completed their exact EOF/hash receipts, totaling 5,304,439 bytes; all 64 selected consumers completed. The registration metadata completed separately with no binary-consumer invocation.

The unchanged route was index.html?candidates&no-log&app=return_fire_demo. Original media, registered configuration and guest clock were unchanged. Actual presentation was Win98Renderer / CanvasRenderingContext2D, with no owning GL/D3D renderer. This remains separate from GL/D3D coverage.

Reviewed search and its limits

Read-only decoding of shipped RFIRE.HLP established bunker J/map, A/left selection, H/deploy, field WASD movement, and the distinction between flag buildings, exposed flags and jeep retrieval. The current map's small central markers did not establish an enemy identity. A selected the helicopter bay; H deployed it, followed by a bounded passive wait. Reviewed movement reached the roadside building, western/southern/eastern home coasts and southeast islets. Later captures show open water. Read-only WRL header/tile inspection helped interpret occupied map extents; it did not establish a target or prove which world was active. The final reverse movement over uniform water has no landmark-based displacement claim.

All 29 requests, including ordinary completion, had successful exact ACKs. Each dispatched input used a matching reviewed capture and live identity; bounded holds retained their independent release records. The audit verified 29 capture bindings, 58 PNG CRC/IDAT checks and the original-consumer joins. These transport/capture checks do not turn the unsuccessful search into an objective result. first-objective-failure.json preserves the unmet boundary.

Cleanup and concrete lifecycle failure

Redline1702's final published release and normal stop/watchdog join preceded allocation; Pawn1723 had also published its offline completion without a box. One fresh box, bx_pvye6uwx, was created using UUID d859d5b2-ae4b-434d-95db-a7b7da626974, retained in watchdog-config.json and boat-created.json.

The normal/watchdog processes were armed and observable before allocation, with an immutable 25-minute deadline. However, I launched their supervisor inside a tool session rather than detaching it. That session returned 143, and all three processes disappeared without terminal or join records. The terminating actor is unknown. Independent watchdog survival and successful join are not established. The strict audit fails on missing supervisor-terminal.json; validation.json records passed:false. No replacement runtime or relaxed qualification was attempted.

Chrome closed ordinarily at 17:38:06.354Z, code 0/signal null. The immediate census failed because the Node controller was still alive; its ports were closed and profile/writer lock absent. A later separate read-only observation confirmed both controller and Chrome absent. The initial census failure is preserved rather than rewritten.

All 400 remote artifact references, 343 unique hashes, were retrieved and independently reread for size/SHA-256 before stop. Since the delegate had died before making any stop API attempt, I executed the explicitly delegated normal stop directly. stop_55f2286b5ccd0f58 completed at 17:40:07.350Z, forced=false. Final API state: archived, archiveAfter:null; expiresAt was not returned. Initial archiveAfter was 17:54:19.014Z. Local lifecycle processes are absent, but no successful watchdog join is claimed. No box reuse/resume/delete, local Chrome, second allocation, public deployment or source/media archive occurred.

Next requires a separately assigned task: establish a readable enemy-target route and a supervisor that survives its tool session before another runtime. This worker stops with objective and lifecycle qualification failed.