Once you achieve any kind of scale, whoever your client is querying to get the book data for those kinds of queries is going to block you

You know that the whole of wikidata can be copied with just a few hundreds of GBs, right? There are plenty of examples of community-driven data providers (especially in the *arr space), so I can bet that there would be more people setting up RDF data servers (which is mostly read-heavy, public data sharing) than people willing to set up their Mastodon/Lemmy/GoToSocial server - because that involves replicating data from everyone else, dealing with network partitions, etc...

Also, there are countless ways to make this less dependent on any big server, the client could pull specific subsets of the data and cache data locally so the more they are used the less they would need to fetch remote resources.

Think of it like this: a client-first application that understands linked data would be no different than a traditional web browser, but the main difference is that the client would only use json-ld and not HTML.