TheMadScientist - 5:32 am on Dec 15, 2012 (gmt 0)
I don't think I'd try to equate that to mobile specifically, but rather the testing mechanism in general ... Result set type 1 for 100 (or N) queries ... Result set type 2 for 100 (or N) queries ... On/Off ... If there's a relatively constant number or queries per device type for the terms and the site/page(s) are shuffled back and forth between types it's likely they'd end up with an 'essentially the same' number of visits, where visits would increase, imo, is if they were 'defined' as 'flexible' and showed in both sets of results simultaneously, but, to me it really sounds like there's a test of some kind going on and sites are getting 'stuck' in the middle.
Say for instance (I'm making numbers up as an example), when 100 people with a smaller device (say an iPad) visit a site and the 'signal' derived from those users == 'user satisfied', the site hits the 'check for mobile' threshold and is throw into the mobile queries for a period of time or number of queries to determine if it's mobile friendly and to get a good sample it could be 'throw into mobile' during different day parts and/or query intent types On/Off ... On/Off ... On/Off for a while ... Doesn't meet the mobile threshold, back to the regular results ... Hits the 'test for mobile' threshold and it's back to on and off again.
Another thing besides my 'small screen' example for throwing the site into the mobile mix could be the G's mobile bot ... Has anyone who's seeing mobile checked to see if there's any type of relationship between G's mobile bot and the traffic changes?