Try the following search:- site:google.com -com
(hope this specific is OK here!)
Obviously the -com means it shouldn't return any results, but it does return supplemental results.
Strangely, for one URL I looked at in detail , the cache in the supplemental index is more recent (28 Jun 2006) than the cache for the same page in the main index (11 Aug 2005).
They also have their own canonical problems. The page I looked at is duplicated in the main index with www. and non www URLs. However, even using &filter=0 the duplicate page only shows up for some queries and not for others.
Interestingly their home page comes up #2 rather than #1 in a straight site search:- site:google.com
I believe Matt Cutts once said they don't redirect from google.com to www.google.com because it gives him an interesting test case. Presumably this test case proves that massive pagerank will overcome most issues?
Does anyone have any idea what the rest of this tells us?