"Open source" doesn't necessarily mean "practical to maintain." Can an open source organization maintaining a fork of Firefox in their free time in addition to the obligations of their regular job keep up with the hundreds of webcompat issues being filed and over 3,000 open issues on the preview version of Firefox on Android alone?
And there also is a fork of XUL-era Firefox already, called Pale Moon, which various people have declared will soon be unusable with the modern web because there aren't enough people that care about it enough to implement Web Components, to the point where the Pale Moon developers were asking websites to not implement Web Components[1]. Are the developers of those websites going to pay attention to a couple of developers maintaining their own Firefox fork that much less people use than standard Firefox? (And we're talking about Firefox, which itself has a single-digit market share.) That on top of the implications of security exploits being found in such a massive codebase faster than the development team can patch them, given how many features are being added to web standards so quickly, probably with the expectation that companies like Google or Mozilla already have the developer capacity to implement them on a set timeline. There's simply so much source code to maintain, and adding the new features that the modern web is going to require to be usable in a few years on top of maintenance takes a significant amount of effort.
The only way that a program as complicated as Firefox can continue to be improved and used by enough people to be viable is if it gets worked on by people with enough time, and I believe that's only practical if people are spending that time working on Firefox because they're salaried. Yes, that's partially because web standards are constantly expanding in scope and complexity, and I'm not really sure what we can do about the complexity creating a hard dependency on Mozilla continuing to exist because, unlike the source code of Firefox, the forces driving those standards are out of the reach of the average OSS developer.
[1] https://news.ycombinator.com/item?id=22604070