It would nice if they could have a perfect system, but given that so much of it relies on statistics and human beings it will never happen. I don't think this failure is a sign of disturbing slopiness. Since December 2002 there have been 2 questions removed from scoring, including this question, which would mean they have a 99.8% accuracy rate with questions.
I guess I just figure that given they're making so much money off this test, and given that it's so damn important for admissions, they should more carefully review every question such that this doesn't happen.
I don't think this follows from the question being removed. As far as I know they remove questions for one of three reasons: the stats for the question come out markedly different from the stats generated when they tested the question; the question had no right answer; or the question had more than one correct answer. If the stats were not funky, then there was no "better" answer choice so whatever conclusions us 99th percentile takers draw are pretty meaningless. The 99th percentile test-takers probably screwed up en masse if the stats are funky so once again our conclusions might not be too helpful.
Just my two contrarian cents worth.
Clearly, if they created a "correct" answer, and it survived all the initial beta testing and review, then there probably is one answer that is more logically correct than the other choices. The main reason they would throw out the question is if an incorrect answer choice suddenly appeared to be arguably acceptable to some degree. (I'd be surprised if they really threw out a question just because more people than expected got it right or wrong -- after all, their perceptions won't change which is really the best answer.) However, it seeems unlikely that an ambiguous wrong answer would be quite as appropriate an answer choice as the choice that was specificaly written to properly answer the question.
Therefore, I'm pretty certain that the best test takers would tend to choose the correct answer, even if there was another arguably correct answer choice, because the best test takers will probably be able to distinguish between a good answer and a so-so answer, even if both are arguably correct.
I guess the best way to determine this (though imperfect) is to poll everyone here who gets above, say, 163 or 165 on this test. (The 170 pool may be too small, as it is by necessity a small percentage of all takers.) If they predominantly chose one answer, that's probably the one that was intended to be correct (and is probably at least somewhat stronger than the ambiguous alternative.)