I've been lurking for a while, but it seems this is the time for me to post.
I'm a professional audio engineer and system designer with a background in concert and recording audio along with experience in lighting and pro video engineering.
Now the pro audio section is quite a lot different from the HIFI world, (I'm a HIFI guy when not working), but the headphone burn-in topic has come up in discussion between some of my workmates and I in the past.
So about a year ago we had decided to test the whole burn-in topic and bought some basic Sony MDR-v6's and hooked them up to SMAART (an RTA anyalizer). The way we did this was to use two balanced Earthworks M-30 measurement mics, and a Sound Devices computer interface. We used some studio foam on each side, then stuck the mics into it at ear level, then played pink noise/white noise and recorded what we saw in SMAART. We ran them for around 20 hours between measurements and noticed a change ever so slightly over around 80 hours of burn in.
I won't go into exactly what frequencies were effected, however there was a clear change of 1-4 DB at many separate frequencies, which tells me that the drivers must have changed their response over time, along with their sensitivity at different frequencies.
I hope this helps!
Your friendly Audio dude