I don’t understand how the tweeter can be surrounded by such a massive baffle area equivalent to multiples of wavelengths of the higher frequencies and and not suffer horrendous diffraction - no, the gentle curve will not do the job!
The beneficial effects of a round-over just start to appear when the round-over radius is about 1/8 wavelength. A 1/4 wavelength radius is probably the minimum for good edge diffraction control, and with a 1/2 wavelength radius we can expect very good edge diffraction control.
On the Magico M9 I guesstimate the radius to be about 4 inches on either side of the mid/tweet section, which corresponds to 1/2 wavelength at about 1700 Hz and 1/4 wavelength at about 850 Hz. So in the horizontal plane, I think we can expect very good edge diffraction control down to 1700 Hz and good edge diffraction control down to 850 Hz. Imo this is excellent in comparison with most speakers.
The reason diffraction is bad is, it is a delayed and distorted version of the direct sound. The problem is that because it is delayed relative to the direct sound, diffraction is not effectively masked by the direct sound. And it happens to be a type of distortion which becomes increasingly audible and objectionable as the overall SPL goes up. Cabinet edge diffraction can also degrade imaging precision, and baffle width plays a role.
I'm not certain but I think the ear's sensitivity to diffraction is strongest from about 1.5 kHz to about 8 kHz, peaking in the 4 kHz ballpark. And in general the wider a driver's dispersion in this region, the greater its interaction with the cabinet edges.
At least that's my understanding.