The only motivation I can discern from the paper is to reuse a name with a different type/meaning, which seems to run counter to every coding guideline I’ve ever seen.
It really seems like the existing solutions (using a different name, like a1 and a2, or an enclosing scope) are better in every way.