This creates three new models that will hopefully make it more clear what an identified user and anonymous user are. Identified users are informed from actual identify events, where anonymous users are informed by devices who have not been identified.
Markdown comments are neat, selects look good, documentation looks clean.
Couple thoughts:
Would any integration tests for these be worthwhile to author?
Yeah, let's see where invalids are coming from, if they still exist, and if we can drop this filter. I get these nagging "we own this and should learn what's going on" thoughts. Also, as this is a public repo, some clarification, if even necessary, will help project confidence in our services.
This creates three new models that will hopefully make it more clear what an identified user and anonymous user are. Identified users are informed from actual identify events, where anonymous users are informed by devices who have not been identified.