I use one non-academic guideline that helps me decide what and how to test in my code.
The guideline is this: If one day I find this thing not working, or suspect that it’s not working, how would I manually verify it? Then I write some code that does that.
If it’s a web UI’s layout that it’s broken, I would probably open the browser console and inspect the DOM. I can do just that with JavaScript.
If it’s an external service wrapper, for example for Firebase, I would probably execute it in the browser console and check the Firebase UI to see that data changed as expected. I can get about the same effect using the wrapped service’s library.