Dusty,
Nice thought-provoking examples.  I think that it is safe to say that not everyone is going to want to use this broadly-defined concept of Individual.  But it will be there for people who want (or in my case need) to use it.  Those who want to use in in complex ways will have to bear the burden of figuring out how.  A few comments inline:

Dusty wrote:
...
This one could be represented as "Bupleurum" for the Individual instance
representing the sheet, but then I would be inclined to establish two
"child" individuals (semantically related to the "parent" sheet), one each
identified to the two different taxa.

So I picked an easy example. Here's a slightly harder one: http://arctos.database.museum/guid/MVZ:Egg:2355.
How about something like this?  Assign an Individual ID to the bird that build the nest which would be an Occurrence (documentary evidence of the bird that built it).  Assign an Individual ID to the brood-parasitic bird that laid the egg in the nest which would also be an Occurrence.  Use the DwC Resource Relationship terms (http://rs.tdwg.org/dwc/terms/index.htm#ResourceRelationship) to define the parasitism relationship between the two Occurrences.  This would be a good opportunity for John to demonstrate how one does this - I've never been clear exactly how it is supposed to work.
...

> Things that aren't taxonomy at all:
http://arctos.database.museum/guid/UAM:ES:3405

Outside the scope of DwC?

Maybe so, but there it is: http://data.gbif.org/occurrences/242032297/. Excluding that would, I think, force you to exclude things like http://arctos.database.museum/guid/UAM:ES:3359 as well - it's all from the same administrative unit. I don't have or want any control over what Curators enter - any scope-limiting filter will have to happen elsewhere.
People are always going to misapply terms.  I think sending records of rocks and minerals to GBIF as occurrences is an error (out of scope).  So I don't feel any need to explain how to handle something like that. 

The point is simply that these are real data. We won't change them to some approximation of themselves or stuff them into a remarks field somewhere. They'll get more complicated before we're done. Anything that's to be useful to us must acknowledge the realities of collections data.

True enough.  Thanks for the challenge!

Steve
-- 
Steven J. Baskauf, Ph.D., Senior Lecturer
Vanderbilt University Dept. of Biological Sciences

postal mail address:
VU Station B 351634
Nashville, TN  37235-1634,  U.S.A.

delivery address:
2125 Stevenson Center
1161 21st Ave., S.
Nashville, TN 37235

office: 2128 Stevenson Center
phone: (615) 343-4582,  fax: (615) 343-6707
http://bioimages.vanderbilt.edu