You couldn't even put a big green poster on a wall of a remote room and ask the viewer what color it is. In the RV procedures, the viewer scribbles crap on a piece of paper, then someone else evaluates the scribbles and finds reasons why the scribbles match what was viewed. The scribbler doesn't claim to know what he saw, in the times I've read their convoluted, obfuscated procedures.
To properly test them on what they claim they can do, you'd need to test both the viewer and the interpreter. Have eight rooms set up with a distinct object in each. The interpreter could know what's in these eight, and even help in their setup.
Then eight viewers would be assigned one room each, and the interpreter would evaluate each, without knowing who did it or which room that person was assigned, and match each drawing to a specific room. You would also need to scramble the assignments, so that if a viewer would not be able to put the letter A on the paper to refer to the room the interpreter knows as room A. The viewer and interpreter would have to be blinded from the scrambling process. For example, the viewer might be assigned room A, but the interpreter would know it as room 4. This mapping that A=4 has to be hidden from them both.
Getting all eight right would be a chance result of 1/40320, sufficient for a reliable test if everything is controlled for cheating.