Yeah I read that proof as more of a reassurance than an emphasized part of the paper. If the method does not even work on convex problems, it's highly suspiciou

Yeah I read that proof as more of a reassurance than an emphasized part of the paper. If the method does not even work on convex problems, it's highly suspicious as a tool for nonconvex problems.