It is incorrect to say that RDA as a standard does not invest in or is not amenable to the RDF metamodel or to being expressed as linked data. In fact, RDA has already been modeled as RDF, and the entire element set is now available, in a variety of RDF flavors, at http://www.rdaregistry.info and https://github.com/RDARegistry/RDA-Vocabularies (in addition to the RDA value vocabularies, which were already available as SKOS: http://rdvocab.info). And talk about proliferation of predicates! There are 955 RDA properties divided among five classes (WEMI + Agent).
For this very reason, there would seem to be even less justification for BIBFRAME to be bound to RDA. As A. Soroka puts it, BIBFRAME should be free to "act as an interchange and point of commonality amongst different models of description," not primarily a transmission format for RDA.
So, here are my questions: how might the new RDA/RDF vocabularies be leveraged to complement BIBFRAME and ease the burden of having to reinvent the bibliographic wheel? If BIBFRAME were freed from having to support RDA in all its granularity, what would it look like?