Canonical mask scheduling study

This is an archived local prototype, not the default or deployed renderer. The exact pre-experiment local client and ground module are restored. Production game, renderer, 322-source selection, world/saves and gameplay boundaries are unchanged. No native artwork files or source approvals change.

Operation probes isolate primary ground-mask work inside expensive material steps. The prototype schedules existing complete passes and bounded 64-face mask portions, retaining original path/triangle/gradient order and rendering resolution. Cancellation restores nested state and never publishes a partial material. 62 targeted tests pass; 18 nonempty native raster cases are exact. These narrow checks do not establish universal whole-world fidelity or performance.

Two candidate normal runs and two unchanged-renderer controls total 48 canonical desktop/phone Region/Hex captures, including reload, neighbor and return. Controls reproduce baseline variation, including the large slate returned-view difference. All whole-screenshot differences are retained, with no tolerance waiver or blanket exact-scene claim. Each captured source remains SHA-bound and the prototype's actual painted-source witnesses remain available. Candidate timing results:

WOG-A-008324: 1440px slice 22.4ms / budget 12ms; scene max slice: 22.4 > 12 | 390px slice 24.1ms / budget 20ms; scene max slice: 24.1 > 20 WOG-A-008325: 1440px slice 9.7ms / budget 12ms; within this tested context | 390px slice 18.8ms / budget 20ms; within this tested context

The slate context passes its current tested budgets; forest remains over budget at shoreline stages. Earlier whole-pass-only trials failed and were superseded; raw local trials remain retained. Further forest shore-mask/contact timing, stable contextual fidelity, lifecycle, broader-device/world and deployment checks are required. Original renderer controls also retain their measured failures, never relabeled as passes.

Atlas: 8,333 maintained / 8,377 full working records; 30 update paintings; 331 source approvals; 7,443 initial reviews remaining; zero full-goal accepted. Only two existing metadata records gain localRenderPrototype evidence. All 8,333 identities, native hashes and approval objects remain intact. Before/after preview: release-preview.html. Full experiment binding: prototype-context.json. Artifact hashes: evidence-checksums.json. Prepared from verified live 5d55469d6c67f70010930cea3cf5b9b8272b3ccc for the existing private repository and Railway service; preparation alone is not proof of publication.