Joint two-timescale fit (brokkr-ruled) + empty-tail cross-check against the A1
corrected anchor (baseline_pad). Deployed PAD-point decay φ0 ≈ 0.95-0.97 across
3 baselines (lofn/mimir/mask), intercept c≈0 → config (decay_rate=0.05, φ=0.95)
is faithfully applied; the R29 net-0.90 was continuous re-appraisal, not re-push.
Trait-flat confirmed; A/P ratio ~uniform (not the S2-expected 1.9); φ_max rec:
relax to ~0.96. forseti signal-poor (low-baseline small excursion, excluded);
mask_neg stalled on a demo hang (7/8).
vh
875f291a42