My (overly charitable?) interpretation is, "Is Checked C so painful to write that we'd be better off just using a better managed language"
Of course, if it's feasible to refactor existing C code like in the paper, this seems like a good path forward, vs trying to "start over" with whatever other checked language of choice you pick.
Given that we are swimming in transistors, holding on to hardware architectures that are rooted in a simplistic model keeps us from making obvious gains in software by having more capable hardware.
