I know I'm answering an old topic but since I had the same issue it might help somebody else.
In my case I found that one job had a step configured on a default database which I detach and the job was running once a week.
The job itself didn't need the default database to run and because of that it wasn't failing.