Okay weird.
What GC miiiiiiight be doing is AB testing (aka A/B testing), in which different versions of the site run simultaneously. As far as us the site visitors can tell, it's pretty much random who gets which version although GC might internally have some rules about that.
It can be a pretty effective way to test how a site works, for example in terms of which version generates more sales -- but there are also major caveats to that and there are definitely right and wrong ways to do it, in terms of getting useful and accurate results. Usually changes tend to be subtle or only affect one detail at a time, so that the effectiveness of that change can be focused on.
Web developers tend to hate it since it can also be an excuse for a lot of unproductive churn and tail-chasing. The page I linked is very positive about it -- Nielsen Norman Group is an A-list interaction design firm and, well, has something to gain by promoting it, though on the other hand they're going to be very straightforward with their facts. It's not hard to find people in the industry getting their rant on about it too and also bringing solid data.
If my guess is correct, within a few hours or a few days either the "Search Keywords" field will return for everybody or disappear entirely. Right now it's in a quantum state.
