I didn't mean "wrong" in the full sense of the word. What I actually mean is that sharding based on user id will most probably not give/guarantee predictable di

I didn't mean "wrong" in the full sense of the word. What I actually mean is that sharding based on user id will most probably not give/guarantee predictable distributions of the writes generated by users (basically there can be no guarantee that chunk#1 of users will always generate the aprox. same amount of data as chunk#2 of users).