For what it is worth, if we play around with the reviewer role we can eventually get the review to close by making the original reviewer an observer, and making the creator of the review the reviewer, and putting someone else as an author.
It makes a complete mess of the review (ie it looks like someone signed off on code that they created), but it works-ish.
Still not ideal, and still a major problem for us. Doesn't happen on all reviews, just some. Not sure what differences there are, none that we can discern in any case.
Let me know if there is anything else that I can do to help track down this issue,
-ang