This is a failure brought by the overreliance on data and telemetry to show what is "most used," without regard for what is most important.
The problem is that everyone is a member of a different 1%. "Oh, only one percent of users have that monitor setup" "Oh, only one percent of users run that app" "Oh, only one percent of users change that setting in any given month" "Oh, that app only crashes once per hundred uses"
If you do 100 things with your computer every month, and every one of them has a 1/100 chance of crapping out, you're looking at a 64% chance of something crapping out on you over the course of the month
Good point (1 - 0.99 ^ 100 ~= 0.63397), and this also relates to the idea that the higher the number of dimensions (in this case features used by a given user), the more volume there is away from the middle. Our spatial intuitions for distributions in two or three dimensions do not prepare us to handle distributions in 100 dimensions.