Sometimes browsers have bugs that cannot be identified with feature detection, in which case the user agent is all we have to know when to work around those bugs.

I feel like with the modern rate of browser development by the time you have identified and fixed the bug the fix may no longer be relevant

This^

I like Firefox, but it's the only browser to blindly follow the questionable and backwards incompatible parts of the SVG-2 spec... I challenge anyone to identify that SVG-2 <use> is implemented instead of SVG-1 <use> without parsing the useragent string.

The former is backwards incompatible (implemented in Firefox) and will render some nice black shapes if you've previously relied upon target context CSS rules for inheritance that have been supported by SVG-1 for the last decade. AFAIK you can only see this in the output since the inside of the <use> element is a shadow DOM inaccessible to JavaScript, so you cannot determine computed styles. For this reason I wrote an "unpolyfill" [0] that implements some of SVG-1 <use> elements with mutation-observers in JS, albeit in a not 100% correct way. Even so, it still needs to be explicitly enabled, depending one something else to parse the useragent string.

We enjoy far more consistency than a decade ago, but browsers are incredibly complex, and there will no doubt continue to be obscure differences that are difficult or impossible to detect programmatically. Rendering differences most of all.

[0] https://github.com/ThomasBrierley/svg2use-unpolyfill

Heh, user-agents have been like that since the start. Almost all modern browsers start their user-agent with "Mozilla/", so we are back to the beginning in that regard. And naturally it is for the same reasons too!