You can't just assert that a bunch of small things added together will remain small, there is no proof for that. That is the question I ask. The dependance and independence on everything that is measured is so complex and intertwined and a small change here resonates to create small changes everywhere.
late response (Click to show)
I know this is not directed at me and I'm late here, but anyway... Is there much justification for believing that some small things added together might make a significantly larger difference? By "add" I think we mean the combination of different elements or metrics, not some arithmetic addition. I don't mean to dismiss it outright as a possibility, but it doesn't seem like an obvious working assumption to make.
You mention averaging out effects from different tips, insertion depth, one IEM sample vs. another, etc., which I agree with in principle. Actually doing so, or extracting the information we want out of the data when there are a lot of confounding variables of unknown influence, is a different matter. Some caution is warranted because the effects from these kinds of things (at least, as measured by the parameters people look at such as FR, nonlinear distortion) are often significantly greater than any effect due to break-in, if data from various sources is to be believed. And any comparison based on auditory memory of course will have some relatively large degree of uncertainty, which needs to be addressed.
I agree about voice coils not increasing in impedance when cooled. Doesn't make sense, unless I'm missing something about the mechanical and acoustic properties of the system. A wire should definitely have higher resistance at higher temperatures. Anyway, many headphones change impedance when clamped on a head; it's very reasonable that some change in the mechanical structure should be able to make some difference in impedance too. Should be apparent when looking at the equivalent electrical model of a transducer (i.e. that takes into account the mechanical and acoustic parts).