lpv: unmodified records accepted (sanity, expect all)                    count      40 /   40 
lpv: lower c0 of one vertex polynomial by 1/20 (W1)                      rejected   40 /   40 
lpv: raise c0 of one vertex polynomial by 1 (W2)                         rejected   40 /   40 
lpv: shrink polytope by 10% about its centre (P4)                        rejected   40 /   40 
lpv: move one vertex by 1e-6 (P1/P2)                                     rejected   40 /   40 
split nodes: remove one plus piece (coverage)                            rejected   40 /   40 
split nodes: remove one minus piece (coverage)                           rejected   40 /   40 
nested nodes: cut lowered by 1/50, no new pieces (coverage)              rejected   25 /   25 
TC/TCS/TCN cap replaced by n2 < 1 (cap check)                            rejected   60 /   60 
SG: p* lowered to about phi(n1) - 1/100 (phi check)                      rejected    2 /    2 
SG (level 0): cap lowered to 0.99 min(n2, khat) (cap rule)               rejected    2 /    2 
near-sphere pieces: lower end divided by 4 (test fails)                  rejected   25 /   25 
near-sphere pieces: label A_n -> A_{n+1} (type check)                    rejected   25 /   25 
A_n (n >= 5) passing with s = +1, re-checked with s = -1 (must fail)     rejected   25 /   25 
nested polytopes (k >= 3) shrunk by 1/2 (P4)                             rejected   40 /   40 
nested vertex polynomial c0 lowered by 1/20 (W1)                         rejected   40 /   40 
tiling: remove one box, sign +1 (coverage)                               rejected    5 /    5 
tiling: remove one box, sign -1 (coverage)                               rejected    5 /    5 
cells: unmodified accepted (sanity, expect all)                          count      30 /   30 
cells: raise c0 by 1 (V4)                                                rejected   30 /   30 
cells: lower c0 by 1/20 (V1-V3)                                          rejected   30 /   30 
cells: remove one cell (coverage)                                        rejected    1 /    1 
sanity: exact facet-minimum bound > sampled minimum (expect 0)           count       0 /   20 max (sample min - bound)/|h| = 6.15e-04
sanity: mbar(B) < min(1, a2/b) at sampled points of lpv boxes (expect 0) count       0 / 2000 

real	0m8.868s
user	0m8.321s
sys	0m0.544s
