Named graphs as documents or fragids?

I've been reading through documentation on named graphs and most of the
examples have the 4th element in a quad as a document.

>From what I understand it can also be a fragid type URI or a bnode.

I was wondering if there were any views about using a fragid in this
position vs a document.

It seems to me that documents are the typical delivery method for triples
over HTTP and I suspect this is going to play nicely with tooling, and
things like TriG, SPARQL etc.

The fragid as the 4th element in the quad seems fascinating to me too.  In
my head it means "this URI is asserting these triples", which I think is a
very interesting part of the open world assumption.

My question is would using a fragid in this position be in danger of
breaking compatibility with parts of the semantic web, libraries, tooling,
spec etc.

( My planned implementation would be delivering a bunch of triples from
database over an API or SPARQL )