Stronghold passive native control state, 2026-11-05

Implemented passive control samples and native boundary intervals in the existing composed owning-Worker observer. Readiness remains false and host presentation remains unknown. Retained original gameplay was not retried. Evidence lives in scratch/runs/20261105-stronghold-native-control-state/; final-inputs.json records exact Git SHA and existing input hashes.

Shared changes and executable tests started only after ReturnFire0305 691020294, MDK0320 68423e88b, and D3D70330 f1d62b136 FINAL runtime AND file RELEASE ALLBUILDSFINISHED. release-gate.json independently verifies normal prior browser exit, exact process identities absent, ports refused, profiles absent, no heavy processes, both locks free and sufficient disk. Predecessor0325 was not reinserted into the queue.

The task follows the missing evidence in the published fresh setup investigation, commit 1bcfd0de488d7ecbf80dd4eb73904ec1f8c46897. No readiness grant, guard relaxation, screenshot whitelist, production input/guest write, clock/scheduler change or original/game/browser retry was performed. Fixture caller/data writes below are explicit synthetic inputs.

The actual original executable is referenced in place: test/binaries/candidates/stronghold-crusader-demo/installed/App_Executables/Stronghold Crusader multiplayer demo.exe, SHA-256 9024c533095f2227b0105b7e8a9e4cf28ec4fb206b165629787b2c97cf99f13f. New samples cannot backfill the retained command-14 history: its pending preset, feedback and selected-control values remain unknown. Static default data and fixture state are not historical values. Source analysis establishes these separate contracts:

The shared factory contributes events/state/pending intervals to the existing owning observer. It has no separate timer, guest export mutation, hook or cap. Existing first-fault, 50M boundaries, 360s, 2048 records, 1MB, ACK and host rest requirements stay intact. The old standalone preset helper also exposes feedback and suppression in its state samples. No historical runtime pin is amended: a changed observer remains a changed dependency for any historical original-run closure.

The executed fixtures use original PE fragments through the actual serialized production owning attachment and shipped OS Worker on both shipped module variants. Caller/data/sprite metadata are explicitly synthetic. The application positive uses out-of-range argument 4 to exercise native entry, return and pending clear without inventing successful preset configuration. Choice 1 deliberately remains pending at its original UI helper; its queue is not an application or completed frame. Independent unobserved and trace-only arms compare all memory pages, CPU registers and identical slice commands. These tests are synthetic fixture evidence, never original gameplay.

Validation passed on both shipped modules, SHA-256 3859c274df0233659eb10fe4d2784fff7d1553242b4ae2b3bf653e5d2ebdd95b and 9f042b70d7e43d3947415e1de01e7a540b3c7f7fa6c7fbefabdbea17e9633e95:

Results and logs: controls-final.json, controls-final.log, native-preset.log, owning-regression.log, build.log, final-inputs.json. All tests are offline; no Chrome, original retry or box was used. Exact Git SHA, relevant binary diff and new-file patches preserve source recovery. Existing media/modules stay at their recorded paths with hashes, without copies, archives or source snapshots.

Remaining prerequisite: a real same-owner/epoch return of an unsuppressed fade-zero draw, identified ordinary host publication, fresh reviewed pixels and warranted control evidence within the unchanged caps. The retained fourth draw remains pending. New fixture samples cannot repair that retained history. Hover, activation, application, draw completion and host presentation remain distinct; readiness is false and host relation unknown.

Box ID/state/expiry: N/A. Shutdown remains coordinator-owned.