My view on Qms is that it is useful to look at non-linear behaviour of real-world drivers (not TS models!). Mechanical losses have a strong tendency for non-linearity. At low cone excursion, the suspension and other moving parts tend to "stick", where as at high excursion they "slip" more easily. I guess I am trying to say that there is a transition from static friction (stiction) to dynamic friction, which is not exactly breaking news. Since static friction is much higher than dynamic friction, mechanical losses depend strongly on the cone excursion. Small signals tend to get lost in stiction at low cone excursion ("loss of detail"). This is not the case at high cone excursion, where dynamic friction prevails. One way to reduce these mechanical non-linearities is to minimize mechanical losses in the driver. This is why I tend to prefer drivers with high Qms.
The fact that Qms is by definition measured at resonance does not take away from the fact that the underlying cause - friction - is at work as soon as the cone moves. Regardless of frequency.
One thing is for sure: a speaker hás to sound different at low levels compared to high levels.
Yes, I would think so. But an electrical recording straight from the driver terminals is about the cleanest way I can think of to record what the driver is doing. If it is "sticking" would that show up at the terminals? I don't know, but many things do.Couldn't this be swamped by other nonlinearities ?