Media isn't federated. The media should just be referenced with a link to the original source.
Normally, the largest use of disk space is the Activity table. It is stored for six months, and only useful for debugging. Below is the Issue, along with SQL commands to check and purge this debugging table. Let us know if this was the issue
There is no easy way to do this. Each Lemmy instance uses their own autoincrement ID for posts. So post 12345 here might be post 54321 on another instance.
It has been suggested that each post gets a universal ID (UUID is an example). Then the local server could just examine the URL and redirect the user to the local post. This suggestion hasn't gotten any traction. The dev team is more focused on fixing the huge performance issues right now.
But when you first subscribe, you’d expect to be missing old posts
OP didn't expect to be missing old posts, hence his question. I had the same surprising discovery. Not sure how the UX could be improved to convey to the user what is actually happening.
Oh my, you are correct. Images are being federated some of the time.
Like most everything else, the intended behavior isn't documented anywhere.