User agents are useful. However they tend to be abused much more often than effectively used

1. They are useful for working around bugs. You can match the user agent to work around the bugs on known-buggy browser versions. Ideally this would be a handful of specific matches (like Firefox versions 12-14). You can't do feature detection for many bugs because they may only trigger in very specific situations. Ideally this blacklist would only be confirmed entries and manually tested if the new versions have the same problem. (Unfortunately these often end up open-ended because testing each new release for a bug that isn't on the priority list is tedious.)

2. Diagnosing problems. Often times you see that some specific group of user-agents is hammering some API or fails to load a page. It is much easier to track down if this user agent is a precise identifier of the client for which your site doesn't work correctly.

3. Understanding users. For example if you see that a browser you have never heard of is a significant amount of traffic you may want to add it to your testing routine.

But yes, the abuse of if (/Chrome/.test(navigator.userAgent)) { mainCode() } else { untestedFallback() } is a major issue.

It is normal practice for each browser to have its own user-agent, no? But the fact that Google intentionally detected it and used polyfills or straight up invalid JS at the time was insane. A similar spin today is "Your browser is unsupported" you see here and there. When a major platform such as YouTube does it, it is really impactful.

It would never do feature detection, would give lower quality h264 video, etc. Back then, there was really nice third-party application myTube which had made this less of an issue but it was eventually killed through API changes.