Yeah, I think servo had the right idea. The main problem with servo's components is that they're severely underdocumented, which has made it harder than it should be for some components (like webrender) to become widely adopted (some of the other ones like html5ever and cssparser are widely used).
I suppose it depends on what the goal was. If the goal was to end up with widely reusable web browser components, the lack of documentation might've been a problem; but if the goal was to improve Firefox, it seems to have been a smashing success.
It has been good at improving Firefox, but if you look at the wider picture then you see that Firefox has been falling in usage and struggling to keep up with webkit/blink. And IMO a large part of that is because core parts of Firefox Gecko and Spidermonkey are much less widely used (by as many apps/companies) than equivalents like Webkit/Blink/V8/JSC, and this because they are not easily embeddedable and their codebases are harder to work with.
From this perspective, not focussing on documentation and making components usable externally is pretty short sighted.
