Except, in the real world, like say this forum, versus the lab, 'subjects' are audio hobbyists, and they *already claim to hear a difference* between A and B.
"Training" is not necessary in this case. They claim to have self-trained. So we are simply testing someone's status quo claim...and their method of determining audio truth.
I would even argue that quick switching isn't necessary, unless that's how the 'subject' has been doing their sighed A/B.
Let the 'subject' repeat their own "A/B" comparison as they normally do...but double blind (and randomize and level match) it. After all, 'even their wives can hear it', so it should be easy.
(In the real world, this doesn't happen either. In every instance I've seen, when golden ears agree to be DBT'd, they demand all sorts of conditions -- which they think will favor them -- to be met, beyond their normal A/B listening practice.)