Native body clearance: source-bound runtime candidate

The preferred private body-clearance study is now implemented as a narrow local candidate in both the primary working renderer and the isolated production checkout based on commit 204f26b. Production remains on that verified commit. This candidate has not been published or accepted as a completed asset qualification.

The three existing native sources are the Cohesive Ward (WOG-A-003376), Walled Gate Quarter (WOG-A-003377) and broken walls (WOG-A-003312). Their original PNG bytes are preserved. Independently decoded alpha128 portrait boxes are bound to the exact approved source URL, SHA-256, asset identity, native width and anchor. A stale or changed binding disables the additional exclusion. These approximate portrait boxes are not exact silhouettes, ground footprints or simulation clearings.

Only the inspected forest profile receives the additional foreground crown exclusion. The original quiet-core rule and background crown behavior remain. Other profiles retain their earlier behavior and unresolved qualifications. Width, bounds and revision callbacks share one naturally selected source per cell. Canonical source eligibility, occupancy, world geography and gameplay rules are unchanged.

The composition revision now includes selected source identity, native hash, geometry, profile and projected position. The primary renderer's existing fast revision reuse additionally tracks a visibility-context revision, so same-array source or approval edits cannot reuse the old body geometry. The isolated production renderer receives the same narrow context and geometry changes without copying unrelated primary runtime work.

Thirty scoped tests pass. They cover independent native decoding, exact alpha128 bounds, stale binding rejection, same-width source and profile replacement, foreground-only intersection, fast cache invalidation, shared source selection, occupied and cleared snapshot behavior, existing canopy regrowth and source routing. These unit checks do not establish full browser lifecycle or performance acceptance. The existing cache-test fixture now includes the real empty coastal, water and ground-contact key functions required by the current unrelated primary cache code.

Normal Region/Hex desktop/phone QA completed twelve views against the isolated 322-source production candidate in a fresh, separately owned local world. It uses normal source selection and camera controls. No response substitutes the candidate artwork or forces a source. All six real pointer pan/return observations reproduced exact camera coordinates,foreground transforms and composed map PNG hashes;page reconnects preserved their local world instance. All six native Hex source-only replays are exact. Every context sheet was directly inspected. One initial server attempt stopped before browser startup because the isolated checkout had no installed ws; its terminal failure is retained compactly. The new attempt reuses existing workspace dependencies without copying or installing them.

Keep every previous visual failure and native artist identity. Phone control overlap, exact silhouette precision, whole-wall final contribution, neighbor/contact, physical scale, other profiles, state lifecycle and full performance remain open. Cold-view equality, unit snapshots and a source-only replay must not be reported as resolving these broader gates. No complete game-ready approval or production-world appearance is granted.

Evidence: candidate file hashes candidate.json, normal production candidate observations normal-production/, preferred private study. The current goal still contains 5,637 pending qualification records; this implementation addresses a concrete rendering gate for three of them.

Remaining production fidelity gate

Actual production-candidate Hex portrait canvases are 958×884 on the desktop viewport and 374×522 on the phone viewport at devicePixelRatio 2. The preferred current local renderer uses 1916×1768 and 748×1044 respectively. Their six native source-only replays are byte-exact,but the final production Hex scene visibly softens fine roofs and stonework. This is inherited production behavior,not a native source repair or a reason to repaint. Port and verify the bounded foreground pixel-density correction separately before treating this candidate as a final quality release.

The corrected lifecycle harness completed all12views with zero browser errors and six exact pointer-return maps. The earlier selector failure retained three partial desktop Region observations and zero accepted complete samples;its disposable world was retired after terminal proof. Current final witnesses are bounded to three context sheets. The Node VM CPU diagnostic is synthetic and does not grant browser performance acceptance.