I suppose that is why I hedged with the words usually and remote. I suppose, given enough time and the right contingencies, anything is reversible.
I've thought that it would be a very cool experiment to slowly work a population of mice & a population of rats back toward their common ancestors, to the point where they could mate with each other.
There's probably more to it than simply determining the ancestral DNA sequences & engineering them into the rats' genomes. But maybe not.