#2 and #3 are saying the same thing: The powers that be know the system is broken, and don't want to fix it, and therefore want to suppress evidence that shows the brokenness.
But let's entertain the idea and say that you get the trial to happen, and your innocent person was found guilty, or vice versa. What do you do with this information? It doesn't help you because you know the outcome already, but in real trials, you don't know, a priori, exactly what happened (unless you were a witness to something, but even then). So a court messed up. What do you do? Real cases don't operate on some outside observer knowing, in advance, what's going to happen. It's why trials are, at best, a good faith attempt at figuring out what happened and dispensing justice appropriately. Everyone just tries, to the best of their ability (ostensibly; I'm ignoring things like biases and such), to piece together the truth, and render an appropriate judgement.
False convictions have a gargantuan cost in time, money, and human suffering.
Justice is incredibly important. We need to get it right.
(Not that I think that money is the reason this doesn't get done).
If you used this argument against something like TDD (in which a 15% extra upfront cost is more than paid for by a 60-90% reduction in produced bugs, according to the Nagappan paper), it would fall flat.
If you do it enough times you will get a picture of precision and recall, which IMHO would be an important determinant of effectiveness. See: https://en.wikipedia.org/wiki/Precision_and_recall