The problem is if you don't consider a rewrite at some point, you're just as likely to have some other product come along developed elsewhere that isn't held back by the constraints of your own and steal your market share.
I don't doubt a lot of rewrites end up being disasters, because insufficient attention is paid to the actual problems being solved by the warts and bits of chewing gum and baling wire in an existing codebase. And potentially because they're simply not done in a way that confers sufficient advantages over sticking with the existing codebase. But I can think of many cases where I ended up spending longer dealing with the consequences of the poor design of existing code than it would have taken me to rewrite it properly from scratch. Hindsight being 20/20 of course.
Or you can keep business as usual while build a separate high skilled team to do the rewrite. Things are hard to build but easy to copy. Unless you're talking about massive progress like OS/Browser, a few highly competent engineers can copy anything in a better structure within a short amount of time.
