OrganicArt exact observed SOFTWARE route: no admitted motion pair

Run scratch/runs/20261118-oasaver-software-observed-natural-motion/ uses exactly one newly authorized local registered scr_oasaver, OASAVER.SCR /s SOFTWARE original. The unchanged binding guard closes at 15,003 ms with UNKNOWN_CLOSED; the missing SOFTWARE natural-motion pair is not obtained. No later endpoint request, replay, behavior repair or additional original is performed. The accepted WebGL pair is preserved without rerunning it: PRESENT 43 → 101, DRAW_UP 1,117 → 2,636, actual page/guest/render eb239… modules and 88 media files, with reviewed motion in the same cloud-backed scene.

Offline preflight compares the preceding successful 4,108-ms SOFTWARE original.cjs, preparation, actual served graph and options. All 1,386 source pins, all 88 media/DLL hashes, the five published helper hashes, registry app, module bytes and Chrome version/normalized flags match. The original launcher is reused with only run/preparation/module-reference paths and postbinding delays changed from [0] to [1000,2000]; original-route.patch records this. The exact published helper retains startupObservations:true, no early Debugger, 15-second startup, 300-ms stability, no-resize attachment and postbind identity guards. No dimensions are forced and no pacing is added to induce readiness. The requested postbind waits are never reached.

The coherent source pin is published Git 47095dcb6c84178f63dde8bb61e9fd142aa1f4b9, explicitly an older pin, not current main. The existing retained module is eb2395e1c115980dd78dbb646a26f755a4119a2a1e086942e715240e07c34bc6 (2,408,519 bytes), referenced in the preceding bounded-startup run without copying it. PNG construction/roundtrip and module validation precede runtime. Relevant shared working-tree changes are recorded as unapplied binary patches. Git remains the source archive. Published47095 native separate-predicate, ordinary/observed equivalence, focused-test and locked-build proof is reused; the native outcomes match only the tested controls, not identical scheduling or performance. No production code or helper changes are made.

The passive sampler records 80 observations, with request-start timestamps from bind +6 to +14,570 ms. Every sample separately reports completed presenter PRESENT 0, composed true, and one visible Screen Saver owned by process 1000, HWND 65537. Device records remain 2, stream records remain 0. Native execution changes across 68 distinct EIPs; yield moves from 5 to 0 and remaining slice budgets range 0–256. Those budgets are not cumulative instructions. Reads of native getters, presentation and composition are separate, not atomic samples. The initial EIP is 0x747f4a96, the final sampled EIP 0x747d854a. Focus 65537 comes from ordinary owning-worker replies with unknown age; modal HWND, API and loading-completion flag remain unknown. No page exception or bridge worker error is recorded. HTTP/console 404s remain unsuppressed.

The personally reviewed whole lossless 1280×960 original-software/startup-colored-geometry.png shows colored curved/fan-like geometry on a black client, the known lower-right white rectangle and static top fullscreen UI. It is not a black frame, despite the sampled stream presenter count of zero. This diagnostic image is not an admitted frame, actual software-executor count proof, cloud-scene comparison or motion pair. No additional interpretation of that distinction is claimed. The scope closes before deferred actual module inventory or executor snapshots: the served module hash does not establish three actual instantiated module identities.

Thus the prior option difference (startupObservations:false in the previous 15-second timeout) remains evidence only. This observed-route timeout does not identify the historical cause, establish observer causation, or provide a native old-fail/new-pass repair. No conditional additional original is earned.

The runtime uses tools/native-supervise.cjs and a separate finite watchdog with independent free-home sampling. Native and command actual waits are exit 1; both command streams reach actual EOF, browser wait is exit 0, and the watchdog is canceled and actually waited at exit 0. No leak is reported; all recorded owned process IDs are nonlive and the known HTTP listener is absent. Minimum free home is 3,064,553,472 bytes, above 1.5 GB; no RAM gate is used. Final evidence records the board-watcher actual wait separately.

Only the selected named PNG, unique logs, hashes, recovery references and run-specific JavaScript inputs are retained; no source snapshot, media copy, frame series or capture bundle is created. Exact final Git, relevant patches, owned new-file patches, readback validation and cleanup precede result.json, written last. This report is committed and immediately strict pulled/pushed. Box ID null, state NOT_CREATED, expiry null, stop receipt null. No workers, boxes, Python, Wine, deployment or foreign-owner edits. Full27, backend parity, FPS, audio, authentic-logo correctness and ordinary native Quit remain open. Ownership is released and work stops.