Native finite Mire against finer wetland ground

WOG-A-009138 remains the complete 1584×1072 native alpha image, SHA-256 c3111d24e68ce1f76ae729ce24c4de0767e2b4ba84e59debcfc6c9bb703e9939. Its generated original WOG-A-009137 and the original wetland floor WOG-A-003578 are preserved. No artist image was repainted, downscaled or recompressed.

The disposable GONZ-CONTINENTS-01 world supplied canonical target cell 1043199. A fresh read-only connected-engine probe verified all 11 cells in the larger finite footprint as unoccupied wetland with the expected features. The engine snapshot hash stayed identical across that probe; no world command or production-world change was made.

Twelve normal desktop/phone Region/Hex views compare the current fine1.25-span floor alone, the retained finite width 240.66390168325452 and a smaller width 150.41493855203407. Both finite cases draw the entire unchanged native image once per ground surface, with its original aspect ratio, native alpha and no pattern, repetition, mirroring or source crop. The smaller width is a placement study derived from source dimensions and the finer floor span; this arithmetic alone does not prove depicted physical grain.

Direct native inspection shows that the smaller placement reduces the oversized moss and stone marks visible in the retained pose. Prefer that smaller pose for the next private contact study. Both poses still form a darker oval against the floor, with a visible perimeter. Reed contact remains readable, but coherent wetland contact and wetness cues need further work. The Region witnesses show the complete perimeter; the bounded Hex witnesses show central contact and must not be used to approve the entire perimeter. The contact-sheet caption names the underlying floor WOG-A-003578; the finite artwork in the middle/right columns is WOG-A-009138.

All 12 candidate lifecycle comparisons are pixel-exact: four normal Region pointer pan-returns and eight canonical Share-view page reconnects. All four actual Hex foreground comparisons are identical to the no-overlay controls. Eight controlled source-cohort map comparisons have changed pixels outside UI. Main-page WebGL2 is deliberately unavailable in this run, so the native floor uses its actual CPU fallback. This is one source, two sizes and one footprint; it does not certify every state, hardware device or GPU/performance path. Earlier raster-history and translated-bound failures remain in the individual record.

The finite layer uses the retained stable method: one transient transparent CPU display surface, anchored at origin [0,0] and bounded only at its trailing edges, followed by an identity copy. No retired controller files, bulk baselines or QA mirrors were restored. Recorded surface byte counts are allocation estimates, not measured total process/GPU memory or performance approval.

A complete in-memory comparison covers all 1,572,864 pixels of the generated original under integer24pixel padding. Every RGB value remains exact. Only original alpha values at least245 become255, as previously declared; every other source alpha remains exact and the outer padding is transparent. The retained alpha image has 1,009,237 opaque, 192,289 partial-alpha and 496,522 transparent pixels. This proves the declared native preservation operation, not visual contact acceptance.

One native comparison sheet and compact source-bound results are retained. Full maps, RGBA buffers, generated input copies and duplicate previews are not written. All owned browser contexts and the server shut down successfully. No new complete game-ready approval, source admission or public renderer activation is made.

Evidence: visual decision visual-review.json, normal-client results normal/results.json, native witness normal/mire-finite-contact-native.png, connected read-only probe normal/read-only-effective-probe.json, native preservation comparison native-alpha-preservation.json, driver derivation normal/driver-derivation.json and owned shutdown normal/owned-resources-terminal.json.