Tell me the scale so I can find something that matches the frequencies.
It also won't nearly be as sharp since he probably doesn't bury the mic in his concha, but it can easily get you a few dB difference.
That's not how statistical significance works.
You "measured" the mouse countless times. Once or twice it jumped. >99% it behaved linearly. It is reasonable to assume that you can use the mouse as if it were linear. Is it working perfectly? No, but that's not what statistic are supposed to tell us.
If you measured the headphones just as often and it showed treble rolloff only once, would you claim it has treble rolloff? I don't think so.
I'm not asking for 100 tests, just at least 2. If you get similar results you probably did it right. If one is vastly different something went wrong.
Another example: We've got one measurement of these headphones showing treble rolloff. Therefore these headphones always have treble rolloff.
Try to start a car. It doesn't start after 2 seconds. We can therefore conclude it will never start after 2 seconds.
I think you'll agree that what someone would actually do is try again. If they get the same result multiple times they'd conclude it won't start and something is wrong.
You say burn in is mostly in the bass? But you're trying to explain a difference in treble with burn in. I don't understand.
EDIT: I've overlooked something. It might actually be break-in. But not the diaphragm. I still don't believe any manufacturer worth their salt would build a driver that degrades that much that quickly.
It's the ear cushions!
http://rinchoi.blogspot.com/2012/04/introduction-it-is-generally-known-that.html
Everything makes sense now!