If a product changes 20dB in 48 hrs, there are a few problems.
1. The material used in the headphones may be unstable. That much change in that short a time may indicate that the material hasn't reached its final state yet. If some aspect of response showed a 20dB change after 48 Hours of undefined "conditioning", it's possible they will continue to change significantly for a very long time. The problem is, nobody would purposely engineer a product that would change beyond normally accepted tolerance of say +/- 3dB...and even that's a whole lot..for the useful life of the product. That's just poor engineering and materials selection.
2. The conditioning signal may have over-stressed the materials beyond design limits. You could for example condition your headphones in a way that would result in a 90dB change, but that would be damaging them with high power. We could be into the edge of that damaging area.
3. There are only two data points here: the starting condition and the ending condition. Two data points are not even enough to define a trend, much less any useful rate of change information. For a proper evaluation of the effects of conditioning, there should have been periodic tests a regular intervals.
4. There was, as I mentioned before, no "control" to the experiment. No means of verifying that the test or measurement conditions didn't change.
5. There was no mention made of a measurement made after a significant period of "rest" or "cool down" after the conditioning. The entire effect could have been the result of deep heating, and that test would have shown that effect. Perhaps the headphones revert after a period of time.
6. There was mention made of 3Hz and 1Hz changes, which are way out of the design parameters for any headphones. If that's the area where these huge change occurred, they could be indeed related to material fatigue, but a 20dB change at 3Hz is well into the areal if inaudibility.
Just a few thoughts, not really meant to shoot down the test because we haven't seen the results yet. But before I would place a whole lot of faith in the generalized concept of "conditioning", I personally would want to know a whole lot more.
All your considerations are valid. I would have liked to have permanent acces to measuring setup, to verify all your ( and mine as well ) doubts. Just for point 6 - area of change with 20 dB was (approx, give or take a Hz or two ) 30 Hz ( thirty hertz ), not 3 Hz, definitely audible and within design parameters and frequency response specification.
Point 5 is the most interesting to me. No they definitely do not revert back to starting condition - audible beyond any doubt. But I would like to know if they do, by how much and after how much time. That would require permanent acess to measuring setup - and that is $$$ beyond my means. Tyll can justify such ( or similar, he has basically Audio Precision setup, which gets calibrated - you guessed it - with Bruel & Kjaer equipment ) expenditure; he sells headphones in one ( if not the one ) of the most lucrative markets on the planet. Slovenia has 2 million souls give or take a few - even if it had ten times the income of the USA, it would most probably not make economic sense, let alone at an income it has and the crisis it is in.
Simple purpose of the whole thing is to squeeze maximum performance from the existing designs. How many headphones that have extended frequency response in bass with an acceptable distortion do you know ? If the diaphragm is too stiff ( almost all as manufactured are ) for the proper extension in bass, at - 20, -10 or - 6 dB say at 30 Hz it does not matter much - it is all too low and "inaudible". Listerners will usually report of a midbass hump in such cases. There are cases with bass response to be in + dB range relative to midrange, but that is rather rare, particularly in really low frequency range, say below 30 Hz. But they do exist. It would be wrong to burn these in any further, as it would compound the problem.
I remember my introduction to the equalizer - a friend had Phase Linear speakers with subwoofer. That speaker goes quite low. For the good relationship with neighbours, he set the 20 ( or was it 16 ) Hz at - 1 dB. At that setting, I thought the bass not to be significantly better than my incomparably more modest speakers. Set it for flat, that is to say to 0 dB or only 1 dB difference, produced a private earthquake. And neighbours banging at his doors.
If 1 dB at 20 ( 16 ) Hz is that audible ( even if it was at such "lucky" point to make such huge perceived difference ) - imagine 3, 6, 10, 20 dB...
I would have liked headphones ( or anything else ) coming from the manufacturer in a "finished" condition. But burn in in audio is unfortunate fact of life - visit any audio fair/exhibition, ask any distributor/importer, whether it was or not a good idea demonstrating a totally new "whatever" that they barely menaged to get from customs etc in time in order to be possible to demonstrate it at all. Most will, if honest, reply that it would have been better to have a totally new unit with zero play time on static display only. I do not believe in toooooo much burn in time stories - but zero play time components just do not sound right. One can shoot oneself in foot by demonstrating zero playtime component - what good does it make to write later on the website that after say 100 hours it came to itself, when countless visitors to the show heard it waking up from zero ? Usually, second chance to present such a product in its proper light comes at the same show - next year - you can bet that competition can and will in the meantime capitalize on such an error.