Having multiple Raven databases is possible, but only advisable in certain situations.
If each database could potentially be on a different server (as one would assume since you're talking about replicating differently) then each must have its own DocumentStore, which is fairly expensive to set up, but this only should happen once at application startup anyway, and you're talking about 2, not 50.
As Matt mentions in the comments, if you have two databases on the same server, then you can use the same DocumentStore and specify the database name when you open the session.
Each database should be for logically very different things. You won't easily be able to commingle data between the two databases. If a document in one database contained a reference to a document id in the other database, you wouldn't be able to use the Include features to get both of those documents in one round trip - there would essentially be a wall between the databases. Indexes could not span between the databases, for example.
Accessing both databases would require spinning up an IDocumentSession for each, both of which would need to be managed separately. If you're managing your document sessions at an infrastructure level (i.e. one session per HTTP request) then having two complicates things quite a bit.
However, if you have a segmented type of application, this can work quite well. For example, if you had a Users database that provided single sign-on across multiple websites (or areas of a website) then this could be a good fit. On most pages the user info would be essentially read-only (like to display the black bar at the top of Stack Overflow), except for the user management pages.
This could also be common if you're going for an Udi Dahan style SOA application structure where you define service boundaries and each service has its own independent database.