Creating a separate SQL table or graph per date sounds like a cumbersome idea. It makes data management much more complex and precludes all kinds of use cases, in particular analytical queries involving multiple dates.

You can simply rely on the power of indexing and its log(N) scaling law. For JanusGraph, see this (admittedly dense) documentation and note that you can create indices for combinations of properties (in particular date and some other property).