Here's another take for actually placing labels on the map (and not out at the boundary): https://patents.google.com/patent/US10102669B2/. (Disclosure, I'm one of the authors of this approach.)

The patent is kind of hard to read, but the basic idea is to choose a number of labels in a region (ie. a map-tile) by the relative density of labels in that region to surrounding ones and then sample them according to some ranking. This preserves relative density so that more populated areas visually have more labels. This works for data too, not just map labels.