If you've allowed your code to grow so far that refactoring is impossible, you have bigger problems to fix than a messy codebase.

Everyone has to refactor into new frameworks later on. Going without a framework early on doesn't add an expense, it actually just subtracts one. It is only if they had chosen the framework (and other code decisions) with perfect future vision that this cost would be mitigated.

I estimate that every 20% that a codebase grows it needs some amount of refactoring. Definitely anything that moves from skunkworks to production will need some pretty heavy rewrites, and will probably need a framework that better meets its newer needs. The question is how much cost was sunk in getting there. The refactor will have to happen no matter what.