I've been focusing on reading, but it looks like there is interest in writing too. The HN API doesn't support writing, so the username and password has to be stored. It's important that reading keeps working even if HN changes their backend in a way that breaks writing. I'll think about it.
It's more about the consistency of operating on data local to the user. For example, see this comment referencing how HN paginates threads at 250 comments for performance reasons: https://news.ycombinator.com/item?id=22231055 A local database does not have that issue.
There have been many HN related websites posted over the years, but a lot end up as dead links. A self-hosted version does not depend on a third party. Another reason is to minimize the round trip latency of contacting a central server. Consider users without a good connection to the server.
2) My focus is on reading, not writing. Local favorites make sense. Maybe with importing public favorites. A user can set their name without logging in.
I'm working on a tool that maintains an up to date archive of Hacker News and a companion tool to browse and search it. I'm not ready to release yet, but here is a dump of the stories, comments, and users from the Firebase API as a SQLite database with a full text search index: https://archive.org/details/hackernews-2017-05-18.db
