A problem that has a really low number of accepted solutions or attempted solutions is going to be harder naturally. Should leetcode rerate difficulty by some sort submissions # controlled for age of problem or give problems an elo and adjust difficulty based on the ratings of contestants that solved it in the contest it appeared in? The latter is probably better because the former can underrate the difficulty of a problem that happens to be popular.

I feel like a lot of difficulty tags as of now are inaccurate. And there is a large range of difficulty between problems of the same tag.