Arctic raster attribution — November 5, 2026

Five individual Atlas records now distinguish preserved painting from an unresolved rendering discrepancy: the full finite arctic ground study WOG-A-008379, limestone WOG-A-003298, standing stones WOG-A-003436, fractured ice WOG-A-003446 and windspurs WOG-A-003447. Original bytes, existing individual approvals, source contracts and next qualification actions remain unchanged. This is evidence for further qualification, not a new game-ready decision.

Two separate four-view normal-client desktop Region/Hex diagnostics each complete three exact lifecycle maps in the disposable local world. The full-method observer first proves zero pixel changes across 1,655,808 native fixture pixels. In the normal client, both cohorts have the same ten operations: one clear, one transform, six image draws, one save and one restore. Context attributes and the dimension-reset boundary match, with no truncated trace or opaque arguments. That cohort's foreground pair is exact and matches the historical A raster hash.

The minimal draw-input collector reproduces the historical 24,610-pixel foreground difference, with 7,905 alpha differences. All four submitted PNG responses are exact SHA-256 matches to the catalog. Post-capture native RGBA hashes match between cohorts across 6,291,904 pixels per cohort. Reading those inputs leaves the previously captured foreground unchanged. This does not prove the underlying image representation at the original draw instant, identify a GPU backend or establish a causal fix. The differing native bounds map to CSS x=211.5–507 and y=802.5–919.5 in the 1440 × 1000 viewport, inside the recorded clipping rectangles. Complete UI occlusion has not been proved; do not dismiss the mismatch as off-screen.

The complete native context sheet is byte-identical to the already retained full-size witness ../arctic-foreground-stability-20261104/normal/arctic-fullsize-foreground-context.png. Both newer runs now refer to that same witness. Seven redundant hints, partial results and duplicate PNG copies were retired after exact preservation checks, reclaiming 5,066,412 bytes. Final observations, failures, lifecycle comparisons, hashes and source provenance remain intact. No raw readback file or new artist asset was created.

Keep the preferred native alpha-domain study and both original paintings. Investigate a bounded native-canvas representation and controlled raster history next; any private candidate must preserve native input dimensions and decoded pixels, retain glyph/other-source fidelity, and pass normal-client lifecycle and full cost checks before activation. Boundary/contact, biome/role/state, GPU, performance and production-world admission gates remain open. Both Citadel hamlets, geography and persistent saves are preserved.

The complete library stays at 9,212 catalog files and 9,237 ledger records, with 5,640 pending qualifications and 3,597 explicit dispositions. These findings grant zero source admissions or complete game-ready approvals. The existing Railway release is monitored independently; its success and live Atlas checks are required before closeout.

Evidence: native-input analysis analysis.json, full-operation analysis ../arctic-raster-operations-20261104/analysis.json, individual findings visual-review.json, preservation registration registration.json, retention plan temporary-retirement-plan.json, and completed cleanup temporary-retirement.json.