Another characteristic of the data which would make random particularly bad is if reloading/recomputing the most popular items took significantly longer than re

Another characteristic of the data which would make random particularly bad is if reloading/recomputing the most popular items took significantly longer than reloading unpopular items. This could be the case for instance if item data size grew in proportion to popularity.