TheMadScientist - 9:58 am on Nov 30, 2012 (gmt 0) [edited by: TheMadScientist at 9:59 am (utc) on Nov 30, 2012]
If it is testing related, then it would confirm the theory G has pre-made datasets of BUYER vs. NON BUYER.
I haven't made it all the way through your post, because I really need to get some sleep, but they already do way more than that ... There's definitely query type separation and caching in use, which, since Caffeine can be updated in near real time, but, the short version is: There's definitely more than only Buyer v Non-Buyer for query types.
[edited by: TheMadScientist at 9:59 am (utc) on Nov 30, 2012]