I'm currently part of George Washington University's BIBFRAME implementation testbed group (http://www.loc.gov/bibframe/implementation/register.html) and I can assure you that there are some of us who are actively working to try and get archival objects described correctly in this environment.
There are a lot of unresolved issues with BIBFRAME right now, so it's very much a work in progress.
There's a lot of promise here - specifically the ability to link individually described materials to the collection from which they draw their provenance - but a lot of the issue seems to be centered around how to classify archival collections since many institutions arrange and describe them differently. There has been some talk of a "Super-Work" or "Collection" level that exists above the Work level in the BF model, or alternatively a Collection as a specific type of the Work class, but we're not convinced that adding additional levels is the right move (and many are actively against it, since that would undo some of the move from the WEMI model to the W-I model).
There are also some interesting parallel possibilities with EAC-CPF and the Authority model of Bibframe, though to my knowledge those connections have not yet been fully explored.
My personal feeling is that we might need to sunset EAD and move to a different description standard that better integrates with linked data, but obviously that's a massive effort. EAD always seemed to me to be, frankly, the way to electronically re-create the paper finding aid. Maybe that's being unfair, but we can do so much more with our data. We need to be thinking less "how can I make my finding aids work in a world of linked data?" and more "how can I get my data into this network?" The finding aid is just one way to do it.
I'm happy to answer questions about BF (as best I can) and bring suggestions to both our testbed and to the larger implementation group.
Special Collections Research Center, Gelman Library
2130 H. Street NW, Washington DC 20052202-994-3925[log in to unmask]