Native ground fixed GPU backing preview

The preferred native1.25hex grass,wetland and meadow calibration remains private. CPU/phase profiling attributes about98ms of the measured desktop cold switch to repeated working-canvas resizes,while texture upload/bind takes about0.1ms. Edge tiles repeatedly changed the backing dimensions. The preview retains one maximum bounded tile backing,changes the active viewport and copies its bottom-left WebGL rectangle to the exact tile destination. The shader,kernel,source pixels,physical output density and native tile plan stay unchanged. The existing2,097,152working-pixel budget is retained;idle backing can now remain at that bound instead of the previous small final edge tile. Full aggregate memory/hardware qualification remains open.

WOG-A-003576: Private fixed-GPU-backing candidate preserves all four whole-map Region/Hex desktop/phone images byte-for-byte at the natural grasslands WOG-A-003301 site,including two exact native Hex subject replays and two exact pointer-return/reconnect maps. It retains the preferred1.25hex grain and native artist bytes. Twelve independent GPU comparisons match28207710pixels;real context-loss recovery matches3387488CPU-reference pixels without partial publication. A separate four-page grassland-cottage timing sample improves desktop cold Hex344→222ms and removes the measured149ms long task;that sample is not per-asset or full-hardware performance acceptance. Keep contact,material/state,broader backgrounds,full lifecycle/performance and earlier unexplained composite gates open. No production activation or game-ready approval.

WOG-A-003578: Private fixed-GPU-backing candidate preserves all four whole-map Region/Hex desktop/phone images byte-for-byte at the natural wetlands WOG-A-003315 site,including two exact native Hex subject replays and two exact pointer-return/reconnect maps. It retains the preferred1.25hex grain and native artist bytes. Twelve independent GPU comparisons match28207710pixels;real context-loss recovery matches3387488CPU-reference pixels without partial publication. A separate four-page grassland-cottage timing sample improves desktop cold Hex344→222ms and removes the measured149ms long task;that sample is not per-asset or full-hardware performance acceptance. Keep contact,material/state,broader backgrounds,full lifecycle/performance and earlier unexplained composite gates open. No production activation or game-ready approval.

WOG-A-003579: Private fixed-GPU-backing candidate preserves all four whole-map Region/Hex desktop/phone images byte-for-byte at the natural fields WOG-A-003307 site,including two exact native Hex subject replays and two exact pointer-return/reconnect maps. It retains the preferred1.25hex grain and native artist bytes. Twelve independent GPU comparisons match28207710pixels;real context-loss recovery matches3387488CPU-reference pixels without partial publication. A separate four-page grassland-cottage timing sample improves desktop cold Hex344→222ms and removes the measured149ms long task;that sample is not per-asset or full-hardware performance acceptance. Keep contact,material/state,broader backgrounds,full lifecycle/performance and earlier unexplained composite gates open. No production activation or game-ready approval.

All three existing four-pane sheets were directly viewed again and verified byte-identical from the fresh normal-control captures. No new PNG was kept. Twelve whole-map PNG hashes,crops,foreground pixels and transforms match the preferred control. Six real pointer-return/reconnect checks also return exact maps. The owned canonical disposable QA world is reused serially after terminal ownership proofs;no production world or gameplay command is involved.

A real WebGL context loss after one completed GPU tile discards it and matches the independent original CPU kernel across3,387,488pixels with no premature output. An abandoned GPU job also publishes nothing. Twelve complete GPU comparisons cover28,207,710pixels at desktop,phone,odd and partial-edge extents with zero changes. Seven CPU material contract tests pass. The separate uninstrumented four-page timing comparison at one grassland cottage improves desktop cold Hex344→222ms and the measured149ms long task is absent;phone cold Hex134→115ms. These are scoped workstation results,not full performance or state-lifecycle certification.

Earlier nine same-viewport composite differences at other recorded cases,full CPU-fallback responsiveness,state changes,other backgrounds/biomes,contact and material fitness remain unresolved. No artist file,approval,source contract or production renderer is changed. Source and actual-world appearance are distinct:these frames are private QA,with zero new production appearance observations.