Problem really simple.
Use harsh criterion of koeff between max and min distances
stable for rounding ang if N=7 also number of pairs with min koeff.
Most difficult situation if N=2 and distance near
50*sqrt(2), if using rounding rooles here maybe very many tricky tests but authors kindly didn't use this opportunities.

Yep, problem rather simple, but i made more then 30 submission before AC, and fully rewrited my program 3 times.
Also there are something strage with the precision.

Yes, there are strange things with precision. After normalizing max distance to 1, not all metrics worked:

sum of absolute values of differences between corresponding distances -> ac

sum of squares of differences between corresponding distances -> wa8

maximum of absolute values of differences between corresponding distances -> wa8

I made exactly 30 submissions before AC. I think this number is standard for this problem. :-)

I think that float numbers troubles tipical thing.
For float math lows don't satisfies and result depend on
algorithm when for right math result dosn't depend on and we can use different algos. It's humor then men play with epsilon  trying to ajust to tests list.