I wish you well with FDW. I thought postgres_fdw seemed great and tried using it to construct cross-database queries. Unfortunately it performed like a complete dog for any kind of batch DML or mildly complex batch select statement, despite me trying every possible trick I could find online. Apparently it's a known drawback with postgres_fdw, supposedly fixed in later versions (which my organisation wasn't ready to upgrade to). Finding much pertinent information about that particular problem either in PG docs or online was a challenge in itself.
It turned out to be far quicker, if less than ideal, to have a program with a connection to each database act as middleman because the queries it ran were locally optimised and could be tuned properly.
All this may, of course, be a sign of a major gap in my Postgres-fu and I accept all derisory replies with appropriate humility.
Obviously there's a lot more to learn about many of these areas to really make the most of them, but this is a really good launchpad.