And I keep saying it will end up flat no matter what. The problem is the time and effort to get there, so if you want to see it happen you have to work with quickly abrading surfaces and reasonable definition of flatness. This is not a limitation though, only practical means to make it observable.
I never specified a time scale in my statement as it is irrelevant - it's about you what the asymptotic result is not about how long it takes to get there - but I think it is observable with our hones, our requirements for flatness and within a time somebody like us may be willing to spend (we sometimes read about people spending hours to hone a razor).
With specialized equipment that you can measure small changes you could see that rubbing two curved surfaces together the curvature moves towards more flat not towards more curved and you'll know where you end based on the direction you're going in.
However, even with a purely thought experiment if increasing curvature is what happens considering the long-long-long term progress should give you pause as you'll be creating more and more curved hones towards something that resembles piece of a marble:
Attachment 201489
At some point you will get down to a small enough radius that you clearly know is going to be flattened by rubbing with another piece.