Almost certainly because no controlled listening comparisons were used. If you look at the TPA3255 datasheet layout examples you will see:
It's almost certain that Fosi have followed this, or the measured performance would be worse. So there is already 'superfast' capacitance right next to the chip where it is more effective than the mod proposed. The mod probably made no difference, which would have been shown by before and after measurements, or recordings that could have been checked for changes.
Don't know where these are, but probably no difference - again no before and after measurements to demonstrate it's doing something useful. It's not impossible that there's a difference (IIRC it came up when opamp-rolling on the 3E integrated showed a difference, traced to a missing cap and the difference in PSRR between opamps) but with the PSRR of modern opamps it would be unusual.
Again no before and after. I don't know the cap specs or the operating conditions - see the
capacitor distortion thread for pointers on how this can effect performance.