OOPS when submitting whether a review is useful

[Michael Nelson]

We (recently?) seem to be getting quite a number of 500 oops in our daily reports when users submit whether they think a review is useful or not (api.handlers.SubmitUsefulnessHandler.create()). The oops reports themselves don't have a traceback [1].

URL: /reviews/api/1.0/reviews/10636/recommendations/
Eg. oopses: 2045rangpur90219, 2045rangpur90221, 2045rangpur90223, 2045rangpur90515, 2045rangpur90517

On 2011-08-07 we had 18 such exceptions, 12 of which were for the one review (id=10636), so it could be something to do with the review itself.

[1] Achuni: will empty oopses be fixed next deploy? AFAICS, it was 181: Anthony Lenton 2011-04-21 that updated the hide_meta option, and preflight tells me we're running version '11.01' of rnr on production which is from 2011-02-03 - ie. it matches, but seems very old... especially given that we seem to have other releases in our PPA for 11.02, 11.03 and now 11.07?)