To summarize the thread so far...
There are currently three different issues:
1) 0.5 V spike at mains switch-on on all outputsAt mains switch-on there is a short 0.3-0.5 V peak on the (unloaded) outputs. Not an issue.
2) 2.6 V (old hardware) / 1.0 V (newer hardware) for ~6 seconds at mains switch-off on all outputsAt mains-switch off, all outputs go to zero an then rapidly ramp up to 2.3 V (old hardware) / 1.0 V (newer hardware) and stay there for about 6 seconds. When shorting the sense terminals to the outputs the voltage increases to 2.6 V and 1.6 V respectively.
This happens even when the outputs were disabled before switching off.
The current is small (1-3 mA), so when the outputs are loaded the voltage is much lower.
This behavior is regrettable, but the voltage on newer hardware is low enough to be not a cause for alarm for most circuits. The 2.3 V on the older hardware is debatable. This could potentially damage low voltage devices, but the voltage rapidly drops with load.
The difference in voltage between old (before 2017-2018) and newer hardware (from 2017-2018) is due to the different output resistance: (1x) 10 kOhm for the old hardware and 1.87 kOhm (4x 7.5 kOhm parallel) for the newer. In theory it might be possible to replace the output resistors of the old hardware to get the same behavior as the newer hardware, but I don't think you should have to make such mods to an expensive high-end power supply of a renowned brand.
3) Massive >10 V spike on output 1 and lesser spikes on the other outputs at switch-offThis happens only when the outputs were enabled at the moment of mains switch-off. This is the scenario that Rohde & Schwarz warn against in their FAQ: Turn off the outputs before switching off the power supply. Under normal circumstances you should not face it, but it is a realistic scenario in the case of a power failure or absent-mindedness.
At mains switch-off, right before the voltage drops to zero (and then ramps up again - previous issue) the outputs show a 40-50 ms peak of over 10 V on output 1 and up to 5 V on the other outputs (unloaded).
This is bad, very bad, and could easily damage circuits hooked up to the power supply. This should never occur in a quality power supply.
Some sources quote spikes of 'only' 6-8 V. That might be due to load or perhaps other hardware changes we are not yet aware of.
I don't know why the voltage spike is lower the further away from output 1 (
see my earlier post), but as a precaution it might be a good idea to connect sensitive electronics to the last outputs.
According to rf-messkopf, pdenisowski and messages on mikrocontroller.net, this issue is fixed in the newest hardware. It is unclear when exactly the change happened. Dave's unit is presumably from 2018 and shows the issue, and so does an early 2019 unit on mikrocontroller.net.