Native arctic domain study — November 4, 2026

WOG-A-009281 is the preferred finite arctic study for further qualification. Its 1536 × 1024 PNG preserves every RGB value of original WOG-A-008379 across all 1,572,864 pixels. Only the alpha footprint changes, using the previously declared polygon and nine-native-pixel feather. All 372,569 opaque core pixels remain exact. Both original paintings and both Citadel hamlets remain preserved.

The sixteen-view private normal-client comparison favors the smaller aligned placement: quiet snow meets the existing landmark apron, blue ice sits beside it, and the full rectangular boundary disappears. This establishes a scoped preference, not full boundary, biome, role or state acceptance. Eighteen tested domain lifecycle maps are exact. A separate export test has four actual PNG candidate views and four controls in desktop/portrait-phone Region/Hex, with six exact candidate lifecycle maps. These use a disposable local world; there are no new production-world appearance frames.

The retained final is tile-library/atlas-v2/readiness/20261104/arctic-finite-native-domain-v1.png, SHA-256 81aec34c21b691da4339e296af80a8b82fbcb57cb553f061f3d4350956afee74. The full original retains its RGB, dimensions and bytes. The new alpha domain is baked into the derivative; a future integration must not apply the inherited domain mask a second time.

Direct drawing from an HTML image failed all eight recorded geometry comparisons against the runtime-mask canvas. Decoding the PNG once into a native-size canvas restores exact native RGBA and all eight replay comparisons. This representation requirement accompanies any future integration. It is not permission to resize, blur, repeat or repaint the source.

Foreground qualification remains open. The desktop limestone comparison changes 24,610 native foreground pixels despite matching recorded image draws, geometry and styles. A stable-capture check and the original full-size retest retain the failure. Isolated context-preference and 48-state warmup/repaint diagnostics reproduce one historical raster hash but do not establish the cause of the other. No speculative public renderer fix is activated.

The Atlas now contains 9,212 catalog files and the full ledger contains 9,237 records, including 5,640 pending qualifications. The 3,597 explicit dispositions are unchanged. This adds one preferred study and three individual source-bound findings, with zero complete game-ready approvals or source admissions. The existing production renderer and 322 source contracts remain outside this publication's activation scope. Broader boundary/contact, biome/role/state, GPU, total cost, lifecycle and live-world admission checks are still required.

Evidence: native visual decision visual-review.json, registration and preservation registration.json, native RGB/export proof native-export-proof.json, representation parity export-source-parity.json, normal PNG client comparison png-normal/results.json, and foreground diagnosis foreground-finding.json. Keep bounded native context witnesses, hashes and compact final findings; no raw RGBA readbacks or duplicate temporary PNGs are retained.