Thing is, by the time you have to take back this application for maintenance, chances are high that the then mainstream framework will be outdated, newcomers won't even know it, the doc will have vanished in a poor "let's move to this new shiny TLD", and still the business logic of the application will be vastly undocumented as well.
I was on your side for a long time, but I've come to appreciate a minimalist, simple, well documented in-house framework over a hype-driven "solve-all-the-thing" one, even free and with a currently big community.
> I've long been skeptical of this line of thinking emanating from major thick client JS framework authors. I think they underestimate what people are capable of
Sometimes even when you can, you shouldn't. Any reasonably-sized organisation will have churn; new developers will have to ramp-up and it will take longer to do so on a hand-coded framework with all it's idiosyncrasies. I'll take Ember over Random Jim's framework, even though he is technically brilliant.