feat(staging): pancake-vs-VR plinth calibration sweep (#225 E1.4e)

[bradleypmartin]

Sub-issue of #225 — fourth and final PR in the E1.4 sub-epic implementation chain under v1.0 staging epic #221. Follows PR3 (panel-backed readout visual).

Scope

Smoke-driven calibration sweep of the plinth's per-scene anchorWorldXYZ, workingSurfaceWidth, and workingSurfaceHeight (and possibly tilt) against both pancake and VR form factors. Likely 1–2 commits adjusting first-pass constants from PR1/PR2/PR3.

Approach

1. VR smoke (Quest 3S, Cloudflare PR-preview URL): each scene at default user pose — does the plinth feel reachable + comfortable? Is the working surface at usable height for sliders/presets? Does the surface tilt feel like a drafting table or like office equipment? Per the sub-epic plan, this is also where the obelisk alternative (the §3.2 escape hatch from PR1's drafting-table) gets revisited if the drafting-table verdict is negative.

2. Pancake smoke (desktop, default OrbitControls pose): plinth + math object framed together on first paint; no manual orbit required. Mouse hit targets work for every interactive primitive. ≥1080p one-third-viewport readout legibility (screen-share proxy per the v1.0 master plan's pancake quality bar).

3. Single per-scene anchor across both modes. The sub-epic plan locks one shared anchorWorldXYZ per scene for both pancake and VR (no per-mode anchors). If smoke shows pancake-vs-VR shared positioning fails, pancake wins the calibration — pancake is the v1.0 audition surface; VR adjusts. File a follow-up only if per-mode anchors actually become necessary.

Likely smoke-tunable knobs per scene

• anchorWorldXYZ X/Y/Z (plinth's floor footprint)
• workingSurfaceWidth (X-extent of the tilted top)
• workingSurfaceHeight (along-tilted-face extent)
• Plinth tilt angle (default ~20°; per-scene override possible if smoke shows different feel needed)
• GRAB_RADIUS_MULTIPLIER_PLINTH if VR controller comfort falsifies pancake's 1.5× value (binary-search bracket [1.25, 2.0] documented in clusterRackTokens.ts)

Acceptance

• Quest 3S 72Hz hold on all four scenes.
• Pancake ≥60 fps maintainer machine on all four scenes.
• VR ergonomic comfort verdict per scene: in-headset, the plinth's position + surface feel like "I can use this," not "I have to reach for that."
• Pancake first-paint framing holds without manual orbit on all four scenes.
• No plinth ↔ stage geometry intersection (StageFloor / StageRailing / StageInnerRailing / ContrastPit) in any scene from default camera view.
• First-pass calibration constants updated in code; doc-comments reflect the smoke verdict on each constant; binary-search brackets + next-step rules preserved.

Possible follow-ups (filed only if surfaced at smoke)

• Per-mode (pancake vs VR) plinth anchors — if single-anchor compromise breaks one form factor.
• GRAB_RADIUS_MULTIPLIER_PLINTH retune — if PR3's value-locked-from-PR1 needs to change after the full cluster + panels are on plinth.
• Obelisk silhouette — if the drafting-table console doesn't survive in-headset evaluation.

Closes #225

PR4 acceptance completes E1.4 sub-epic. Closes #225 (all five children a/b/c/d/e shipped).

References

• Parent sub-epic: feat(staging): control plinth hosting interactive UI (sub-epic) #225. Epic: feat(staging): production-quality environmental presence (v1.0 epic) #221.
• Sibling PRs: PR1 (E1.4a/b + quadrics), PR2 (cluster sweep + readout slot registration), PR3 (panel-backed readout visual).

🤖 Generated with Claude Code