Sorry for commenting again but I think there is a way for you to do this in a completely open, easy, and privacy-preserving way. You don’t need to access their database.
Get a list of instances that you want to look at the subscription patterns for. (All the instances here + Lemmy.world)
Go to that instance’s website, click the “Communities” tab at the top, and then click “All” It shows how many users from that instance are subscribed to that community (both communities from that instance and outside of that instance)
If you find a way to automatically (or manually) scrape this data from all of those websites you can create the visualization that you were talking about.
So you were right, the data is open source it is just specific to each website.