I'd like to support Casey's point about whole-part relationships not being all of one kind. In the dim mists of the past, MARBI was considering a proposal to support the creation of relationships between separate items that had been bound together and so necessarily shared a barcode. (I think this was it: http://www.loc.gov/marc/marbi/1999/99-02.html). It's interesting how similar the discussion seems to be to some of the FRBR relationship discussions.
In any case, the discussion of the proposal ultimately came down to the differences between bibliographic relationships (between editions, for instance) and physical relationships (items bound and circulated together, sometimes for obvious reasons--issues of a serial--or for reasons unclear but relating to practices a century or so ago). I was dealing with miscellaneous collections of pamphlets bound together at the time, so the problem was not abstract from my point of view.
As I recall, the much-to-simple solution proposed originally by a vendor were not accepted, largely because of the discussion at the meeting. Although these situations seem so rare that they should not interfere with our quest for simplicity, in fact they are often cases that can teach us quite a bit about the limits of simplification in whatever model of the world we're looking at.