Owned native limestone preparation — November 5, 2026

The preferred private limestone representation now has one explicit native owner. WOG-A-003298 keeps its original PNG and exact decoded 1586 × 992 pixels. The owner checks its validated source descriptor on every use, reuses a backing for a new image object with the same source key, releases it on replacement/revocation/pressure and permits at most one native backing. WOG-A-008379 remains the unchanged full-size ground context; its rectangular boundary, large ice grain and snow-contact concerns are not accepted by this work.

The real-image fixture passes 80 checks, including 32 replacement/pressure cycles, revocation/regrant, invalid dimensions, missing fingerprints, budget rejection, permission errors, allocation failure/retry and idempotent disposal. Its maximum declared native backing is 6,293,248 bytes, with zero remaining after final disposal. These are backing-accounting values, not total browser/GPU memory measurements.

Two eight-view normal-client runs cover desktop 1440 × 1000 and portrait-phone 390 × 844 Region/Hex views. All twelve regular lifecycle comparisons and eight authorized foreground comparisons against earlier references are exact. Four post-capture owner probes preserve native RGBA and the captured foreground while exercising image-object replacement, pressure and revocation/regrant.

The second run connects the owner to the application's actual image LRU and approved-source checks in the private native-rendering harness. Four controlled LRU trims remove the real cached record, release its prepared backing and reload a different image object. Four approval-list revocations remove the source from painting. Four fingerprint changes keep path permission true while the owner rejects the stale cached source. The three remaining image draw identities, geometry and styles stay exact in each suppressed state. All twelve restored foreground comparisons are exact, and the original source contracts, version and 160-entry limit are restored. Browser-only test copies never alter the source/world files, geography, production contracts or saves.

Native backing is reserved before allocation and counted alongside source-image pixels against the existing cache model. Tested source-cache usage is 163,575,692 bytes on desktop and 138,412,652 bytes in the phone layout, under the host's 268,435,456-byte tier. The test does not qualify 128/192 MiB tiers, total process/GPU memory, physical-phone performance or total renderer cost. The harness retains its existing private native backing/material preparation; this is not an unmodified production-runtime performance benchmark. No public renderer activation follows from these results.

Every one of the sixteen native observation crops exactly matches the already retained desktop context ../arctic-foreground-stability-20261104/normal/arctic-fullsize-foreground-context.png or phone context ../arctic-native-canvas-phone-20261105/normal/arctic-native-canvas-context.png. Their existing native comparison witnesses remain authoritative. Six redundant combined sheets and completed-run hints were retired after exact preservation checks, reclaiming 9,591,558 bytes. Complete observations, failures, source hashes and lifecycle findings remain; no raw readback files are kept.

Continue lower memory tiers, complete renderer cost and pressure/state scope, clean production integration and combined testing with the preferred alpha-domain study WOG-A-009281. Keep the original HTML-path discrepancy and older phone hash as historical counterexamples. The full catalog remains 9,212 files and 9,237 ledger rows, including 5,640 pending qualifications and 3,597 explicit dispositions. There is no new artist asset, approval, source admission or production-world appearance claim. Both Citadel hamlets and every original/final painting remain preserved.

Evidence: actual client analysis analysis.json, owner analysis ../arctic-native-canvas-owner-20261105/analysis.json, 80-check native fixture ../arctic-native-canvas-owner-20261105/owner-fixture.json, individual findings visual-review.json, registration registration.json, retention mapping temporary-retirement-plan.json, and completed cleanup temporary-retirement.json.