I totally agree about the small size and comfort of Xelentos and IE900. Other than dynamic range (which is helpful only if you're a stage musician needing faster ways to destroy your hearing), single driver headphone tech seems to be just as competitive these days and has found methods to tune FR which are every bit as flexible as those in the large multi-driver monstrosities.
The 'technical capability' argument is pretty fascinating, but probably confused by the fact that people might not agree on what they mean by the term. We've spent a couple of years looking into this, measuring the total waveform error from a range of headphones (including the Gen 1 Xelento and IE600 - but not IE900 yet, sorry) once corrected for whatever tonality the headphone possesses. In other words, once EQ corrected, 'technical performance' is the total error from whatever remains, from whatever source, including THD, IMD, the (in)ability to respond to fast transients, phase, decay, slew rate/transient response of the driver and/or amp, RMF noise from the cable, etc. We call this total non-tonal error (or NTE). The measurements are actually quite stable in that they don't have quite as large a variability as FR does with things like eartip type and insertion depth. But a word of caution - there's still a lot we don't understand here. Also, there's a valid criticism of this type of time-domain measurement - not all errors are equally audible. Some errors (e.g., phase rotation/distortion) might be completely inaudible. So take all this with a large pinch of salt
That being said, it's still intrinsically interesting to see the total waveform error from the perspective of a mathematician. If your dog poops on the floor, but you don't see it because it's behind the sofa, you might still want to know about it, and in an ideal world, you'd still hope your dog wouldn't have shat on the floor to begin with, because it's hard to be 100% sure there's never going to be a future circumstance in which you're going put your foot in it.
Technical performance seems to have some contributing factors we would have expected (BA drivers having inherently higher harmonic distortion than dynamic drivers), but also some unexpected factors. It seems that low NTE favors IEMs that don't have wild swings in their FR curves - in particular it favors IEMs that were flatter to begin with, regardless of EQ correction. As I said, there's still a lot we don't know, but there might be a plausible explanation for this, which is that every sound any human has ever heard has already been distorted to some degree, because our external anatomy 'messes up' the amplitude and phase before the sound ever arrives at our eardrums. We need some of that 'mess' for things to sound normal to us, because our brains have adapted to our anatomies. That amplitude correction is already accounted for by comparing FR against a target that includes the effect of things like the 3-4 kHz pinna gain, but that means some level of accompanying NTE may be inevitable.
So whether 'technical performance' is something that is actually useful, i.e., could provide an additional metric (on top of FR) that would be helpful in assessing headphone quality or value-for-money is still up for debate. Sean Olive thinks definitely not. I suspect not (other than obvious outliers). But that won't stop certain reviewers (whose ears & brains certainly aren't capable of accurately assessing total waveform error) from continuing to supply all us mere mortals with a grade from D- to A+
There's a longer rant on technical performance
here and some initial NTE measurement results
here. Lower errors have the histogram peaks farther left; higher errors peak father right. The histograms are basically bins of summed error (in decibels) over 400 ms windows. What's remarkable about these curves is they're statistically converged from actual music test tracks - play any broad range of music and over a long-enough period of time and the headphone error converges on the same NTE curve.
