I have been testing the performance of Fedora objects that link to many other repository objects (for example a collection with many members, or a book that contains many pages).
I've written up some of my findings in the Fedora wiki. But the TL;DR version is that it's reasonably fast for small numbers of members, but gets progressively slower for larger numbers of members. And the move from LevelDB to a relational database for storing metadata looks like it's adding to this issue. So a book with 500+ pages or a collection that contains thousands of objects slows down to 30+ seconds to read the RDF from Fedora.
I mostly wanted to let people know about this issue so it can be a factor in discussions about modeling, performance, etc.