When we get our unit test infrastructure in better shape, it seems we 
should have a test like this:

crudeSift xs == fancySift xs if xs only contains crude siftable things

fancySift doesn't exist right now (except maybe as helpers to 
siftForPending).  It's also not clear to me yet if fancySift can be 
defined as something that doesn't use crudeSift (it seems like it 
should)