ftp://bryston.com/pub/reviews/Swedish Review 14B SST, part III.pdf
I mentioned earlier that a French publication tests amplifiers in series. I was mistaken it was a Swedish publication. The above is about a Bryston amp they tested. Bryston actually changed the output filter design which made the amp go from not passing the test as accurate to passing it.
Their method of the bypass test is to load the amplifier under test with a difficult dummy load with reactances to match a real speaker. They then tap the output, divide the voltage down to unity (input matches output within .05 db or less). Play the amp (thru another amp) while comparing amp in circuit vs amp bypassed. A group of people listen to and describe the differences they hear. Then they continue the comparison blind.
They report two amps have been so good in the sighted portion of the bypass test nothing was noticed. One of those two was nevertheless detected in the blind portion of the test. Only this revised Bryston was undetectable both sighted and blind.
I don't know how many amps they have tested this way. I have done a few myself though only listening sighted in my case. Not many amps are capable of being undetectable this way. Very good ones get rather close. Some well liked amps don't get close leading me to think they are liked for their carefully styled inaccuracies.
I wish this method of amplifier comparing was more common. It would slay a number of entrenched myths.