It might be a bit controversial notion. But when I try to look for extensions say for completion framework there are many available and I used ido and then moved to ivy. I knew helm existed and is much more powerful in terms of feature-set, but never used it.
I read that emacs can report on keyboard usage (I think from xahlee’s posts). Can it be repurposed to monitor which extensions are used often in users setup and they can share that report if they wish to public say via melpa like services. I suppose that melpa downloads can be used as a measure of usage.
When I hear of a package that may be interesting, I immediately check its repo page to see how many issues and PRs are still open. I look to see whether they have garnered any responses, especially if the submissions are of high quality. Years of issues building up isn’t a good sign. This isn’t 100% reliable, as different skilled developers approach issues and PRs quite differently, but it gives you some information. And there are outliers, like multiple-cursors, whose developer is very skilled and motivated, but whose popularity overwhelmed his resources.
For simple package, “no updates” for 5 years is usually fine. But before investing energy in a larger new package, I want to know whether it will still be working well in the next 5 years.