I once was involved in a complete rewrite in a new language that ended up saving the company -- so those do happen, empirically.

That was an unusual scenario, however; the previous iteration of the product had major issues, buggy and costly to maintain, and the language change was in part for the purpose of repointing the hiring pipeline toward a different kind of profile. Such special circumstances aside, I'm with you on recommending against rewrites. Much as I'm often tempted.