Finite native tundra ruin: canonical context qualification

The selected original-RGB matte remains a private proposed canonical ruin source. A complete scan of the unchanged 2,252,800-cell default world finds three natural tundra ruins and 55 arctic ruins. The proposed contract deliberately retains the legacy family's tundra role; arctic availability is not inferred from a snowy painting. Two retained interior tundra candidates pass in-memory natural/effective state and occupancy checks. The first, cell 2027561, has six tundra neighbors, including a farm. Actual server state must be checked again before browser observation.

The 1254-square source has nonzero alpha bounds [314,412,915,832]. A proposed full-source width of 260 world units gives a visible ruin width of about 124.8 units, comparable to existing 126-unit ruin portraits. Anchor [0.49,0.65] places the courtyard/contact near native source y815.1. The full image is drawn once with no source crop, stretch, repeated detail slots or repeated ground paintings. This is a proposed physical scale awaiting normal desktop/phone Region and Hex review, rather than an approved layout.

The published canonical parser compares source URLs literally to permission files and rejected the normalized tile-library permission proposal before any server or world scan. The private browser test therefore serves the same exact native PNG through a source URL under /assets/region-composition/source/, with one matching private contract. This changes no public source file, parser, renderer or permission. Any later public transport/admission decision must be separately recorded against exact native bytes.

The first local server attempt terminated before creating a browser because the isolated published checkout could not resolve ws. Its failure and terminal evidence remain retained. The corrected attempt reuses existing primary dependencies through NODE_PATH; it installs nothing and changes no runtime source. Inspect normal-v2/results.json and its terminal evidence when they exist. A running process or a source record is not a painted context or game-ready approval.

Complete finite family/detail/quiet-ground hierarchy, actual alpha and material contact, native grain, scale, neighbor occlusion, nonrepetition, lifecycle, fallback, full memory/performance and source admission remain open. The artwork adds no functioning fortress, player ownership or defensive action. Every original and selected final remains preserved, as do earlier rejected native-matte and missing Region glyph findings.

A normal selector diagnostic corrects the initial candidate-only preflight: cell 2027561 naturally selects the existing cold round wall from the complete source pool. The new matte is accepted as a private contract, but is not chosen there. Evaluating both retained interior candidates against all 40 private canonical sources finds cell 2035674 naturally selects the candidate, with a nearby village. No source ID, seed, feature, biome or selection code is changed. See selection-preflight.json and normal-v5/initial-binding-diagnostic.json; all earlier zero-frame attempts remain retained. Normal-v6 uses the naturally matching second site.

Four normal private desktop/phone Region/Hex views now visibly retain the complete ruin at the naturally selected second site. One bounded native-pixel witness is retained. Region uses a 512-square display input and differs from native isolated source replay by 31,692 desktop / 31,760 phone pixels, maximum channel difference 255. Hex uses the 1254-square native input and its isolated replay is exact in both layouts. Neither replay is final-pixel attribution, and no edge-only explanation is established. Region fidelity remains an explicit failed gate; full family/detail/quiet-ground, contact, scale/occlusion, lifecycle, physical-device and complete memory/performance qualification remain open. The preferred wall/ground studies were not painted in these four parent frames. No source admission or public renderer change follows. See analysis.json and visual-review.json.