I wasn't going to even bother, but I emailed Richard and got a response.
Hi Sean,
Yeah, we had some issues with scaling up. It seemed like we'd worked
out a pretty good system during early testing, but when we scaled up
the number of HITs our internal reviewers got overwhelmed and the
consistency of their work dropped off. After hearing from a few other
AMT workers we're going to try switching over to a qualification-based
system, so instead of rejecting HITs we'll just be disqualifying
consistently poor performers.
I'll look at your HITs later this morning and double check any
rejected ones. Please let anyone else know that we're up for that.
Really sorry for the hassle.
Rich