Had a look at the google docs.
The measured voltages cannot possibly exist on capacitors.
The peaks you see are caused by leakage of SMPS, one that operates around 0.5MHz.
Of course, even when you modify the crap out of the power supply you will never get it to suppress 500kHz needle pulses anyway.
For this only decoupling caps work.
That's the biatch when using a scope. Most likely when you connect the ground and probe together on both pins of a component (so basically shorted input) you may still see them.
If those voltages were really there the currents would have to be huge and extremely short. In other words the peaks you see are measurement errors. The amplitude changes might as well have come from how the probe and probewire(s) have been draped over the DUT and table.
DAC's and amplifiers, certainly with regs feeding the output stage, differ substantially.
The reg mods in a D30 cannot possibly cause the perceived improvement. It looks more of a subjective case of associating.
Technically I agree that with such large capacitors on the outputs of the regs you would need to install reverse diodes on the regs.
Also that large capacitors on the output of regs is suboptimal.
However, it looks like that it works and larger caps can provide transients much better than LM317/337 regs can.
Output power is drawn from the other 2 large caps though and the 1 Ohm + large caps also form a low pass filter to remove HF noise.
I don't know how many Modi's have been returned to Schiit with blown power supplies. There may be some merit to John's claim that a blown reg (would effectively become short between in and output and thus overvoltage) may well be the root cause of blown output stages.
Could also be SOA being reached of shorted output devices peak currents.
Jason probably knows by now.