suggests aligning bones then applying the recalculation
That still doesn’t explain why the two of us get different results with the same rig.
%<
No it doesn’t, and I sure can’t explain it.
Actually I was looking again at your image and I can’t work out which part of the model it is. I guess it’s the fingers but there seem to be angles that aren’t in mine? My pic was looking diagonally toward his left hand down the arm toward the neck. The fingers are straight, the thumb arched above and the arm essentially straight but not perfectly.
My picture was without IK target bones whereas yours looks like it has IKs. Not that that explains much because I had this trouble even when the IKs were still there and even after I dumped all constraints and even on simple tests I did with just a few loose bones.
What is really weird is that it was so much better when it was rotated just a little then recalculated but broke if recalculated again after straightening…? It sure acts like a bug.