The article I referenced is about 10 years old. I'm quite sure the magnetostrictive sensor technology has evolved significantly since then and that it will find corrosion in embedded rods as long as one end is accessible to fit the sensor to the rod. I am not sure how well the instrument can quantify the metal loss, but I think its fairly good.

I would be very wary of corrosion predictions in concrete, although I have no direct experience with them. It seems you would have to rely upon very good to excellent characterization of the concrete and the environment and hope for a very narrow range of predicted corrosion rates so that when you multiply rate x time you get a meaningful estimate of metal loss. Theoretical models may be OK if you want to do a very coarse sort on which concrete formulation or operating environment might be worse, but I would not want to rely on them as an absolute indication of remaining strength in the rebar.