Cool tutorial! I've seen both the data transfer and shrinwrap modifiers w/ vertex groups used to preserve the shading in situations like this. Is there a reason to prefer one over the other? Looks like data transfer can do a lot more than just this. How about a more general tutorial on the uses of the data transfer modifier? ;-)
Thank you and good question ccocheret !
Using the shrinkwrap modifier will only virtually project the vertices to the reference sphere. In order to get the exact same shading as the reference sphere, those vertices need to be placed exactly at the same spot (as the reference object). This is because the normal of each face is defined by its orientation, hence by every vertex that define it.
So, projecting the shifted vertices on the reference sphere will still not provide you the proper shading. You need to mimic the face normals behavior using the data transfer modifier.
I hope this answers your question!
Regarding your second question, thanks for the suggestion: this modifier offers indeed quite a lot of possibilities :)
/Thibaut.