[AnimalMuppet]

But if it works out badly in practice (enough times, for some value of "enough"), don't you have a basis for saying that the idea doesn't seem to work in practice, no matter how good it sounds in theory?

You don't want to do that after one failed attempt. You probably don't want to do so after two or three. But after enough attempts, you probably have a basis for saying that, no matter how good the theory sounds, it won't work in practice next time either. And if each time it fails produces large amounts of human suffering, there comes a time to quit trying the same stuff that has worked out badly over and over.