Great job. I managed to reproduce that and instrument it.
Here is what the O2 is *backfeeding* into its input when you pull out the power supply:
View attachment 14539
The worst case spike is a little hard to see but I put the cursor on it (dashed horizontal lines) and it 16 volts!!! The amount changes from run to run but it is always many volts.
Given this, the fault is in the O2. The output of the DAC should have not transients fed back to it. This is likely feeding into the power supply rails of the Topping D10, causing the processor to get screwed up. Yes, that is a technical term.
I guess Topping could put mitigation for this also but seems to me this is an O2 problem.