In the collection list, would it be possible to include rollup counts for each section?  For example:
+ Rammstein (3/34)
  + Mutter (11)
  + Reise Reise (11)
  + Sehnsucht (12)

On each level you see (A/B/C) where A is the number of immediate subcategories, B is the number of second level subcats, and C is the number of tracks.  So basically A=sum(B), B=sum(C), C=count(*).  Of course only show as many as makes sense (if you're only grouping 2 levels deep, show A/B).  Another syntax might be (A, B, C), (A-B-C) or (A) (B) (C)

For me, I have a lot of complete albums, often multiple albums by the same artist, but I also have a lot of singles where I have a single song from a single album (or maybe 2 songs from 2 albums) from an artist.  Sometimes I'm ok with an ecclectic mix of whatever happens to come up (so I'll use random/dynamic mode in this case), but sometimes I want to queue up a couple of complete albums.  This would help me visually distinguish what's what much more quickly.

Finally ..I knew I can't be the only one who would find this useful...
I wrote a simple patch for those (song) counts some time ago, but it was rejected mainly due to performance problems. (unfortunately it isnt really in a usable state currently). See and vote here: http://bugs.kde.org/show_bug.cgi?id=123406