By the massive free volume between them. Even if you clamp the two metals together tightly, unless the surfaces are atomically smooth you'll get void spaces that are large kinetic barriers.
Even if you are dealing with atomically smooth surfaces, you'll get a grain boundary which is still a kinetic barrier, and a large grain boundary too.
That being said, your friend's point is still fundamentally valid - why wouldn't an atom on occasion cross the boundary and switch sides?
And yet, vacuum-welding is a long known actual thing, and even without vacuum right in ordinary human environment, by good, non-trivial, but not magical surface prep. But it's probably more than just proximity. In vacuum there is probably also evaporation and a form of electrolosis.
Optically flat isn't even needed per se. You can weld metals by slipping two blocks a bit under pressure side by side just like that in standard atmosphere, and when there's some sort of low-pressure environment it just takes that much less of a slip. No need to go down to space-grade vacuum at all.
Even if you are dealing with atomically smooth surfaces, you'll get a grain boundary which is still a kinetic barrier, and a large grain boundary too.
That being said, your friend's point is still fundamentally valid - why wouldn't an atom on occasion cross the boundary and switch sides?