One ridiculous example I remember (and which tends to surprise people who have a more limited context of the past): MySQL seriously used to have a page of their documentation dedicated to why views were a bad idea and there was a good reason why their database hadn't implemented them. Of course, they eventually realized why people wanted them, so they are both now implemented and the page telling you why you shouldn't use them is gone.
The gotchas listed here (https://sql-info.de/mysql/gotchas.html) have all now been addressed, but I still have a bitter taste from that era and avoid MySQL to this day when I can. I feel the same way about PHP despite the improvements—let's face it—the developers were forced to take. (https://eev.ee/blog/2012/04/09/php-a-fractal-of-bad-design/)
Together they made an unholy alliance whose ecosystem's sole reason for existence seemed to be maximizing the production of high severity CVEs.
All established languages either grow features, or wither and die.
C seems pretty much a counterexample; sure, it's added a few spec changes since ANSI C or C89, but C is essentially stable.
Get ready for K&R compatibility removal on C23, with improved generics, type inference and lambdas probably landing on C26.
C is actually a good example, given that it shares authors with Go, which at the time also decided to ignore what systems programming languages have been doing since 1958 outside Bell Labs.