:-) I don't know that that is what is really happening, Tanel's advice is good for that. The implications I was making were that it should be tested in the latest release, and we can't really know what unpublished bugs are doing or what their patches are really fixing (without some in depth research, anyways). But we can see that various things with analytics are fixed at various times.
Also see Note:285913.1 suggests Alter session set events '10119 trace name context forever, level 12'; of course, that's about an ORA-600, but hey, we're in discovery mode here, trying to figure out if that analytics problem may be related to this one.
kksParseCursor shows you are parsing.
kksLoadChild you are hard parsing (loading a new child cursor to library cache)
apa* functions are hard parse driver functions
kko* functions are ones who are responsible for query compilation, transformations and validations etc and it looks like that's where the loop is happening (if these functions are on top of the stack in most of samples taken).
As I said previously, parse times in minutes mean you've hit a bug, so Oracle support should be able to help you (send them the same stack trace and optimzer environment trace/10053).
If you want to experiment around you could set some optimizer's cost based transformation parameters fals and see if you still hit this bug.
Like for example set "_optimizer_cost_based_transformation"=false and set "_optimizer_push_pred_cost_based"=false .
However I'm pretty sure it's a known bug by now so the best thing to do is send this information to Oracle Support!