== 1m-n4096-t16 repair threads=16  CPU-s: median of both copies of each arm (n legs); A/A = worst |X - X2| / min over the reps; verdict clears the A/A floor or is UNRESOLVED
      m |     fold    force     F/T |   aa_F   aa_T  floor | verdict |      wF      wT  wF/wT | foreign max | peak MB force/fold
    192 |    26.44    43.03   0.614 |   3.9%   2.5%   3.9% | fold    |    2.68    3.56  0.754 | 17% (8 legs) | 4208/2098
    256 |    29.52    43.38   0.680 |   1.5%   2.1%   2.1% | fold    |    3.08    3.60  0.855 | 19% (8 legs) | 4214/2098
    288 |    31.35    43.55   0.720 |   1.3%   1.4%   1.4% | fold    |    3.27    3.62  0.904 | 16% (8 legs) | 4218/2098
    320 |    32.03    43.32   0.739 |   1.4%   2.3%   2.3% | fold    |    3.42    3.65  0.936 | 52% (8 legs) | 4222/2097
    352 |    33.45    44.20   0.757 |   1.8%   0.4%   1.8% | fold    |    3.57    3.67  0.972 | 19% (8 legs) | 4221/2097
    384 |    34.92    44.48   0.785 |   0.5%   0.7%   0.7% | fold    |    3.76    3.71  1.014 | 19% (8 legs) | 4226/2108
    416 |    36.89    44.62   0.827 |   3.0%   2.5%   3.0% | fold    |    3.90    3.73  1.044 | 31% (8 legs) | 4229/2098
    448 |    37.81    45.41   0.833 |   3.2%   1.4%   3.2% | fold    |    4.09    3.80  1.078 | 23% (8 legs) | 4232/2108
    512 |    41.59    46.65   0.892 |   2.9%   4.9%   4.9% | fold    |    4.50    3.89  1.157 | 108% (8 legs) | 4240/2097
    640 |    47.34    49.83   0.950 |   2.8%   1.5%   2.8% | fold    |    5.15    4.10  1.254 | 31% (8 legs) | 4255/2097
  crossover (log-interpolated, median): CPU m ~ >640   wall m ~ 373
  foreign CPU over 80 legs: median 12% of a core, max 108% (after-leg median 13%)