You did the experiment I asked for and then some, thank you.
One caveat before anyone reads too much into it: you changed two things at once, the snapshot mode AND the replica power state, so the "fell back to a full" on the two ADDS VMs cannot be pinned on the forced shutdown by itself, right?
The line I would look at in your screenshot is "failed to open disk" under Clean VM directory, since that step works on the existing chain rather than on the new transfer, and it fits a fallback to full better than the snapshot mode does.
@tjkreidl's SMlog question is the right next step and I would follow that rather than anything from me.
I still cannot tell you what the power-state change did to the chain.
