In my experience, that's easy to solve from an operations point of view and it just take a couple of easy to teach tricks to avoid it.
However, the scaling limitations of traditional RDBMS are insurmountable when trying to do things like data science, for instance.