"Figured it was a bad idea" actually means that some people were against it because they believed semantic class names were the solution, I was one of them. This was purely ideological, it wasn't based on practical experience because everyone knew maintaining CSS was a bitch. Heck, starting a new project with the semantic CSS approach was a bitch because if you didn't spend 2 months planning ahead you'd end up with soup that was turning sour before it ever left the stove.

Bootstrap and the likes were born out of the issues the semantic approach had, and their success and numbers are a testimony to how real the issue was, and I say this as someone who never used and despised bootstrap. Maintaining semantic CSS was hard, starting was hard, the only thing that approach had going for it was this idea that you were using CSS the way it was meant to be used, it had nothing to do with the practicality. Sure, your html becomes prettier to look at, but what good is that when your clean html is just hiding the monstrosity of your CSS file? Your clean html was supposed to be beneficial to the developer experience, but it never succeeded in doing that.