Native edge repair: original retained

A built-in imagegen edge donor was integrated through the approved original-preserving masked method: only existing partial native alpha could decrease; every original RGB value, all 1,009,237 opaque native pixels and existing transparent pixels remain exact. The 1584x1072 candidate changed 130,340 partial-alpha pixels without resampling. Actual Chromium native decode preserves the opaque core exactly. Twelve normal desktop/portrait-phone/landscape-phone Region/Hex views use identical painters with only the source PNG changed; all 18 comparison lifecycle maps are exact and three Hex foreground comparisons stay unchanged. Direct native/context inspection does not show a clear ground-transition advantage; retain original WOG-A-009138 as preferred. The donor and proposal are unselected working inputs eligible for retirement after final checks, not new catalog assets or approvals. Preserve their hashes, method and final context witness. Normal private context is now explicitly recorded in the canonical backlog, while visual role/contact, relevant GPU/state/performance and source-admission gates remain open.

The source is the unchanged native alpha image, SHA-256 c3111d24e68ce1f76ae729ce24c4de0767e2b4ba84e59debcfc6c9bb703e9939. The unselected proposal is 050334345878666febe8d5a3be6e9c50cd3eabca8c6fe85aebd41412a5695c7e and its generated donor is 60ee146f40902f072902274aa73c929c99bd079ae32cd4455c90fe7ae3d5ef9a. Built-in image_gen supplied only donor alpha to a bounded masked integration. No generated RGB was copied, no source file was overwritten, and integer24pixel alignment preserved native dimensions. This separates intentional native-edge alpha edits from the prior display-alpha operator; interpolated world-edge pixels can change even though every opaque native input pixel remains exact.

The comparison uses the same smaller authored pose, seed and read-only11cell wetland footprint as the prior retained study. Both versions draw one complete native PNG with no pattern, mirroring or crop. Main-page WebGL2 is unavailable, so this is a CPU context/lifecycle comparison, not broad GPU or performance acceptance. Normal framing follows the admitted reed; source-rectangle containment and native-center geometry remain distinct from complete screen visibility. No camera override is used to force the private mesh into view.

The original remains the selected artwork. Neither the generated donor nor the proposal becomes a library record or a new final output. Their explicit private-input disposition, exact hashes, bounded context witness and preservation results remain after temporary retirement. Every catalogued original, approved reference, final and both Citadel hamlets is preserved.

The canonical backlog now links current retained context evidence and describes private observed placement separately from an admitted source contract. Completed native-input/CPU evidence is recorded with its scope; broader visual/contact, GPU/state/performance and admission work remains pending. The old no-context-frame placement explanation is superseded without approving any profile or changing live appearance status.

Evidence: visual outcome visual-review.json, normal source-pair results normal/results.json, native witness normal/mire-native-edge-context.png, native integration proof masked-integration.json, Chromium input-core proof browser-core-preservation.json, source/edit prompt and provenance plan.json, owned shutdown normal/owned-resources-terminal.json and retained original.