There is one simple reason that enables all the shit mentioned in other replies: as much as we like to pretend, the impact of design is very hard to measure.
A product experience is nuanced, part of a journey, with learning curves, preferences and problems. Much of this is completely ignored in the measurement of the success of design changes.
There are lots of reasons people tend to screw with things that are fine, but the root cause is that they get away with it. Hell, they even get rewarded for it.
This seems like a solved problem as well. Don't people use focus groups anymore? Or have an open feedback forum where people can submit feedback and use user ratings to see which issues are the most important to the most users.
Many companies already do have forums like this and they're filled with UI complaints having sometimes thousands of upvotes, but it's all for nothing because almost none of them actually use that feedback.
My bet is still on "UI design isn't a full-time job, it's project work, so UI designers have to keep moving things around to stay relevant in one company". As much as I don't like outsourcing, having a bunch of designers on your payroll with nothing to do most days but still needing to pass all the employee rating crap is bound to result in an unstable and never good enough UI.
A product experience is nuanced, part of a journey, with learning curves, preferences and problems. Much of this is completely ignored in the measurement of the success of design changes.
There are lots of reasons people tend to screw with things that are fine, but the root cause is that they get away with it. Hell, they even get rewarded for it.