I took one for the team and had my HMP4040 serviced by Rohde & Schwarz (Service Circular 16282). Price... 1111 EUR without VAT or shipping. Ouch. It might be a bit cheaper if you can deal with R&S Service directly, but I had to go through a partner (
Esenwein GmbH - very friendly people). The turnaround time was two weeks.
The power supply came back with a new calibration certificate and a elaborate 12 page test report (it was not re-adjusted, but still well within specs).
I've tested the power supply and can confirm the issue with the >10 V spike is fixed.
Measurements:
All outputs enabled and outputting 1 V before switching off with the mains power switch. The profile is identical to that from before the fix apart from the faster ramp-up to 1 V after switching off the power supply (compare against my screenshot in
post #11). Thus the issue with the outputs staying at 1 V for 6 seconds after switch-off remains (but we know that is present in the latest revisions as well and it's not a big deal). The faster ramp-up seems to suggest some capacitance was changed:
Let's zoom in at the point of switch off. That's what we are most interested in. Gone are the massive >10 V spike on channel 1 and the 4-5 V spikes on the other channels. There's still a small spike of almost 2 V on channel 1 and 1.5 V on the other channels, but that is acceptable. Let's not forget that under normal conditions you should not run into this issue at all. It's only during a power failure or absent-mindedness that the power supply may be switched off without turning off the outputs first. This will not damage any devices under test:
Now with the voltage at 2 V before pressing the mains power switch:
The profile of channels 2-4 is exactly the same to that of the screenshot posted by rf-messkopf in
post #33. This suggests my HMP4040 now behaves the same as the newer revisions. In fact, I'm quite sure the hardware modification applied to my power supply is exactly the same as what the newer revisions come with from the factory.