I agree. I had the feeling when it was released years ago that the whole v4 API was kinda generated from a given database Schema. It was very usefull if you wanted to fetch data in a similar way than how the website does it. e.g. pull requests are part of repositories. I for instance wanted to query all pull requests for a given org and filter based on repositories. The search api was in most cases the only solution or go and mix and match with the v3 API.
I remember something from the Changelog podcast that Github has a number of single- or two-person projects that are experiments but are then used by a non-insignificant amount of people. So then they choose to keep it on life support with minimal resource expenditure.
I agree. Graphql API is basically a database schema. That thing is lifting GH from heavy burden and will mislead their consumers like us.