Hate to rain on everyones parade, but a 0.9 THD is not a very clean power source, I would not want to hear this system with middle of the road speakers and certainly not high efficency speakers. 
I suspect that this was posted as flame-bait, but anyway.....(and this will hopefully get to Code Chemist's question as well)
First, taking any single measurement as an authoratative indication of sound quality is laughable, *particularly* for a mass-market product where spec sheets are usually somewhat fictional. In the case of the Panasonic units, it's clear that they *really* wanted to get the 100w/ch number, so they just threw out whatever the number happened to be.
I don't see any lit on a TAS5036/5182 setup as in the XR25/45, but the docs for the TAS5076/5182 EVM are enlightening. The 5076 performs a bit better than the 5036, but the trends should be similar.
The primary observation is that the distortion at low power is significantly lower than at full power - like all amps the distortion skyrockets when you start to hit your headroom limits. The distortion numbers for moderate levels (<10W say) are around 0.015% to 0.025%. It rises to 0.1% at high power (95+W) before hitting the wall at 100W at which point the curve goes more or less vertical.
The panny units might be slightly higher than this, but there's no reason to think they'll be substantially different at moderate levels - maybe 2-3dB worse. So, it really isn't at all out of line with other amps/receivers in terms of the numbers. It would be very interesting to compare the spectral balance of the harmonics to see whether there is anything interesting there.
In terms of the source of distortion, the biggest non-changeable sources are probably jitter on the master modulator clock (since it's derived via PLL from the I2S master clock) and the output bridge drivers/mosfets - the timing on the outputs is probably the main source of higher-harmonic distortion in the non-feedback Equibit process.
The two biggest addressable contributors will be the power supply and the inductors in the output reconstruction filter. Waynes mods probably help a bit on the PS front by reducing the ESR of the supply. I don't believe that Wayne does anything on the output filter side, though.
You could probably do a whole lot more on the PS front with more extensive mods, but there isn't much room inside the case and you have to be careful with layout due to the EMI from the switching noise in the vicinity of the output stages. Adding a carefully selected bypass cap or two the the 'lytic at each bridge output to give a good low ESR over the cmoplete audio band would be great. Running it off batteries would be even better.
Bypassing the output inductors altogether would be a great way to significantly reduce distortion - particularly IMD - but you'd be dumping a ton of 386kHz switching noise down your speaker cable which is not a stellar idea. For anyone willing to do major surgery, putting the amp into a larger case and using high-guage air-core inductors would be an interesting mod, but you'd have to figure out how to shield them. I believe Tact does something like this in their Equibit amps.
