You ca interpolate the values from the original. The efect isnt 100% the same. But then the endtire idea of using history would be wasted.
So let me explain this a bit further. A meshsmooth gets interpolated over the points next to it and the points and the points beyond that so each point in a smooth is affected by atleast 4 points in each direction. Now if you bind this to a skincluster and smooth then it goes ahead and proceeds working like this as long as history is intact.
BUt if you broke the correlation and rebound by interpolating the weights... youd end up with something that would behave differently. Why, because your points now only relate to the binding stucture! Whereas the soothbind provioiusle related to how the parent points were now mowed to by the cluster, and proceed still working like previousley described.
But would it be possible to mimic the effect precicely, Yes! Absolutely almost identicaly too. BUT and a big BUT you couldnt aeffact how the skin is bound! The skin would only do this under certain very spesific circumstances. Namely where the original skin is bound in a very specific way so that each point is affected by a number of joints that correspond to how the smooth interpolation is done. Ad this wold make certain bone configurations impossible too. Wich means it would be sortoff useless.
So it can not be realy done well, altough the interpolation is still probably wat you all are ACTUALY looking for.