I'm a great believer of that premise too, but my latest discovery on my pair of big Sony TAN77es power amps has me scratching my head.
They are 430 apart in serial numbers, all the components are identical. Both are multivoltage units with no EU filtering. Both pull the same current from the wall and are running secondary rails within a volt or two of each other.
One is an ex Sony dealers demonstration unit, the other came from a Sony State manager's own collection.
They are both in excellent condition. Both test identically on my bench. Ruler flat to over 300kHz, distortion is below what I can measure (<0.001%) and power outputs are identical into 8 ohms. 4 ohm power output was difficult to measure due to my dummy loads getting close to destruction.
Intermodulation plots reveal nothing between them.
With a 100Hz, 3kHz, 10kHz square wave, their reproduction was identical. (I use a square wave to sort out my amps in short order).
So it isn't transient, rise, fall, power, response in the bottom or top ocataves, power bandwidth or anything I can test.
They sound different. I swear. The ex demo unit sounds better than the ex Sony Sate Manager's unit. The demo unit had a lot more hours on it.
I will try to record the difference by padding down the tapped off output into a DAT recorder and posting the files (when I get a chance)
I have subsequently tried several other pairs of identical amps (yes I have way too much gear) and subtle differences exist. (bear in mind I trim the flatness of each amp using a square wave and the tone controls where a tone bypass doesn't exist- this is very important in older amps as their response is usually far from flat when bass and treble are in the flat position)
If amps (or receivers) sound different, but measure the same, you're measuring the wrong things. I've never seen an amp that didn't reveal the reason for its "sound" on the test bench.