Sorting by population is very common. Unfortunately, most variants won't have a population record. Valencian (ca-valencia) is one example.

For writing systems we estimate the population using locales and/or languages. We can do something similar for variant tags.

It may be good to add 2 figures

1. cited population from locales
2. population upper bound: derived from the population of prefixes/languages associated with the variant, eg. roh_rumgr could have as many users as roh in general.