The flipside is that performing joins in a process across many document databases (which is what this approach would resemble) is wildly non-performant compared

The flipside is that performing joins in a process across many document databases (which is what this approach would resemble) is wildly non-performant compared to doing it in a relational database.