So if your code is spaghetti and that makes it impossible to add new features (happens easy, especially if you let an LLM loose on it), the refactoring to fix it would never happen and the engineers who stay or are willing to be hired will just be those clueless enough not to know they won’t be able to achieve anything.
That certainly happens. On the other hand, engineers are typically way too eager to claim a spaghetti code. We need to rewrite the whole thing. After long experience, I've concluded most cases of people complaining that something needs to be written because of spaghetti code are really just somebody other than me wrote it and I disagree with some design decisions without understanding why those decisions were made in the first place and I think I can do better. I was involved with the very large rewrite some years ago. The code is now in production for 10 years, and so we call it successful. On the other hand, most of what we learned along the way of doing the rewrite was why the original code was designed that way in the first place, and so a lot of design decisions that we thought we were going to be able to rethink were designed that way for good reason in the first place. Even the places where we did make a big improvement because of what we still believe was a bad design decision in the first place. Not only could we have refactored in place to the new design but also the new design I can now say in hindsight was also a terrible idea. We're now trying to figure how to refactor in place to a third option which only time will tell if this is a good idea but we know for a fact two different ways to design this part that are not a good idea.
I'm skeptical of your claim that an LLM is really going to add more Spaghetti for Code. Now, certainly, especially the LLMs of last year did a terrible job of designing code. However, they have gotten much better just in the course of one year. But even at that, most of the problem, I think, is you are not taking the time to understand what the LLM is doing in the first place. An LLM is by definition code that was not written by you and therefore it is a lot of boring and tedious effort to understand it. However, if you take the time to actually read the code, it often is perfectly understandable and just as good as code any other human right. But that other human is also not you and so if another human wrote it you would also call it spaghetti code.
Good part of LMS is they are very good at refactoring code. Whether they wrote it or a human wrote it, they can do a lot of changes for you that would be tedious or take a long time to do by hand.