Hm... I made a checking script and let it eat some FBM's contained in DAZ's M4.
Tried with "FBMHeavy" and "FBMBeerbelly".
And indeed, the children have exactly the same morph delta values on the shared vertices than their parents. Unless my check code is wrong, but I hope not.
So my conclusion for today is that this is the way to go, and there are limits I have to accept and situations where bad shading has to occur because the morph is so extreme.
Ah well.