It is more about maintaining consistent output. The rail caps provide the current for an immediate need at the output devices. All larger rail caps can do is help with transient response.
The other reason behind changing them is due to the regulation scheme for the amp. As close as I can tell, the rails are regulated based on two things. The actual rail voltage, and the voltage difference between the +40V rail, and the +40V raw supply. If the voltage sags at the rail node, but not at the supply node, then this signifies high current draw, and a low impedance load, and the power supply will start ramping back the rail voltage until the raw and rail are equal. This is what allows the amp to put out basically the same power into a low impedance load, as a higher impedance load. By adding rail capacitance, you minimize that sag, and theoretically let it drive an increased voltage to a lower impedance load w/o the regulation backing the rails down. This could destroy the original output devices. Since speaker impedance varies w/ frequency, something as simple as this could keep the amps from ever knowing it was driving a low impedance load.
I do however plan to totally eliminate the feedback loop to the PS that drives the low impedance voltage regulation, as I plan to run these amps at 1-2 ohm mono.
M: M100, M44 for a custom amp project
Zx: Zx500, Zx450, Black Zx350
ZxTi: 4 Zx600Ti's, 1 Zx400Ti
Ti: 5 800.1's & 900.7 for a custom amp project. 1 1200.1, 1 1000.2
Tantrum: 2 1200.1's, 1 600.4, 1 500.2