Is it preferable to do load dev with unfired or once-fired brass?

I do it on new brass, and then confirm the same load works on fired brass once it's sized. If your sizing (case body and neck tension) is different than the new brass, then you can get a slight change in the load.

I had made some good groupings with new brass on the 6.5 Creedmore, and next time out with neck sized brass of the same loads shot differently. I believe its the volume of the full sized case is smaller than the volume of the neck sized brass resulting in that variation.

If you full size new brass and fired brass, it may not change so much. Neck size is what changes.

I tend to take new brass and load it then neck size the next 2 firings. After that I start bumping shoulders and I have never seen a difference in the load.

Depends on how much brass you have for the rifle. When I got into 6.5 CM business I ladder tested, and group shot on virgin brass, because I started with 500 pieces. Once they were all shot once I stuck with the same load, never retested, and didn't notice a change.

But if you have 100 pieces of brass for a rifle, it's not too far down the road that you've shot them all once. After a .002" bump there's be nothing wrong with testing groups within the node again, but I wouldnt re-ladder test.