Fr0zkaandClaude Opus 5.5
5389abed2e
Unreal per-tile timeline: per-tile Core work is not slower than headless (FI 11.2 vs 16.2 ms per shared key); the gap was tick-paced admission with a small queue (FI ~5.2 of 8 slots busy). voxel.StreamingReadyQueueDepth default 16 -> 64 (effective 32 with 8 slots, total cap 40); result backlog cap 256 unchanged. First admission -> LOD0-2 settled: FI 2.72 -> 2.10 s, SW 2.37 -> 1.31 s (3 of 4 runs; one 6.9 s outlier re-admitting 3 LOD0 keys, cause open). apply_ms max 4.2 ms (14 ms budget). Frame-interval max 118-145 ms both before and after (open). World unchanged. Fast tier PASS; Linux UE host builds. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016dhxM43dnWvUu2B5Ph1iG1