This is quite interesting, because a package which gains a high amount of votes in a day will have a very high popularity for a short period of time, which is a little misleading.
I thought it would be based on the number of people looking at the package/webpage or cloning the repository. It is a little misleading because a package might not be popular but still have a high popularity, however on the other hand it does help new packages gain recognition quickly.
I guess what's misleading is the word "popularity" because it does not properly reflect what it is nor its purpose. As "Hugo Osvaldo Barrera" said in the thread I linked in my initial answer , I think the word "trending" might be more suited as this is clearly what it is. This is the term GitHub is using for the same purpose for instance .
What is considered a "good" popularity rating for a package?
I don't think there's such defined "good" popularity rating. I mean, the higher the better :P
For what it's worth, the 50 most popular AUR packages  have a popularity between 45.51 and 5.85 (at the time I'm writing those lines).
 https://aur-general.archlinux.narkive.com/XuG8fEVO/how-the-popularity-value-...  https://github.com/trending  https://aur.archlinux.org/packages?O=0&SeB=nd&K=&outdated=&S...
Regards, Antiz (Robin C.)