One possible factor is the proliferation of LLMs to write code for us. I noticed that a few versions after Jetbrains implemented LLM integration, the bugs in their products skyrocketed. Also, in my own job, it's often tempting to use LLMs where I really shouldn't, by which I mean where I can't easily check the code to ensure that it's fully and subtly correct. For example, if I'm working with an unfamiliar library I might ask an LLM what the incantation is to do something, and then check it by reading the docs of each thing that it references. I regularly find issues when doing so, but a different developer who didn't check these docs may miss this and ship a subtle bug.