Sparse indexes required to handle columns with large variance between values

[ykulbak]

Good example are timestamps where there's hi chance that the number of index entries will be very close to the number of keys in the region