Alice reserved-world capture attempt and startup readiness repair, 2026-11-03

The new original-game attempt stopped at a host startup race before any guest batch. The first fault-status request reached port8123 before the control server listened. The retained repaired request writer saved the primary ConnectionRefusedError; ordinary supervisor shutdown exited0. No native world1022 capture, movement stack arguments, Player allocation continuity, neutral/airborne comparison or wind/plank/knife progress was obtained. The accepted predecessor cave observation and offline reserved-world repair remain distinct from this new failed attempt.

Exact source input was 11462d25b5729d0ab481b57292654e3a2c624091, including Alice fault retention 0c2bc2da5641695157703a778a377e6f44b16d74. All5909 staged inputs passed full hashes from exact committed Git blobs and separately pinned dependencies; foreign WIP and Jedi changes were excluded. Unchanged compile inputs against accepted module input200832ae permitted reuse of WASM SHA256 0a58fd7b9ac82c5303048776f442b15e72528b55ddf85564e0df92fcb1aa384d. The module observer confirms that actual2089591-byte module was instantiated. No build or accepted prerequisite fixture was repeated. The43 external pins comprise15 original game files (engine plus14 companions),27 pngjs files and actual stdole2.tlb. Local full hashes passed before boat creation; remote full hashes, actual imports/startup mounts/fonts/PNG and original manifest passed before execution. No compilation archive download occurred.

The exact pinned post-stop contact.cjs, capture-repaired.cjs (installed as capture.cjs), request-repaired.py (as request.py), receipt_failure.py and world contract were verified before staging. They are preserved in scratch/runs/20261103-alice-reserved-world-ground-route/, with inherited pins. The prior contact/capture repair remains native unexecuted. The receipt writer did execute on the new transport error and retained it successfully.

First failure and bounded repair

PID9123/session ca3b5662-a918-4f23-9a8b-dfd4273ef94b started 18:37:49.668651Z. Command1 was submitted about7ms later, before control bind, and failed at18:37:49.702671Z. The supervisor's ordinary quit was command2; exit0 was18:37:50.008613Z, elapsed0.339962s. The CLI records zero API calls and zero batches. The request was never resubmitted. No guest memory writes, injected calls, input, config/save edits, skip or cheat occurred.

Task-local startup_ready.py and commands-ready.py repair admission by waiting for the same live PID/start tick and the original server-listening callback line before the first command. They read host files only and send no probes or retry commands. Prior first-error, process exit, absent/reused PID and bounded startup timeout refuse admission. The callback line is emitted inside server.listen's callback in exact lib/control-server.js.

Seven focused controls passed locally and on the same boat: delayed process start/listener, retained failure, exited process with stale log, reused PID, absent PID, timeout without submission and a current live listener. Readiness also refused the actual failed session with prior failure retained, with no new request. Python syntax checks pass. This is a tested host-only repair; it was not followed by a new original-game launch. The original one-launch attempt, first-error and module/lifecycle receipts remain unchanged. No production source changed, so a new full build was unnecessary.

No DLL was loaded far enough to discover its actual session base. There was no held movement phase, selection diagnostic or stack argument observation to save; the complete available launch/module/failure/exit diagnostics are kept. Missing native bytes cannot be reconstructed. usableForFooting=false.

Recovery and shutdown

All5942 remote payloads passed local size/full-SHA verification before delegated normal STOP. Archive114949218 bytes, SHA256 c97cd7d1e0b7fe8807fa70f9a42e6d579475507ab2a3398596c5f015b29baab1. The run retains RECOVERY.md, SHA256SUMS, archive-verification.json, source and dependency pins, complete recovered runtime, primary request failure, module observer and focused readiness tests. All own helper/runtime PIDs were absent and8123 closed before STOP. Local disk stayed above2GB.

One fresh boat bx_pxc2kr3u, TTL3600, expiry 2026-11-03T19:37:21.022Z, was recorded before use. RUNTIME OWN followed Jedi's explicit18:30:57Z release. RUNTIME RELEASE was18:39:24Z. Stop stop_56d1dde137c9b3a6 completed18:39:49.477Z, error null, forced false. Actual state STOPPED at18:39:49.449Z; completed available snapshot verified 18:39:49.384Z. Original expiry is now inactive; custody returns to coordinator. No predecessor/foreign boat, resume, subworker or public deployment was used. Single task ends without a successor.

The remaining gameplay work is unchanged: use repaired readiness before a separately authorized original session, validate the complete world1022 capture, retain actual lifecycle/stack anchors, prove Player allocation continuity, obtain two neutral supported samples and ordinary airborne comparison, then continue the wind/plank/knife route. No later session is launched by this task.