Project_ids presumably don't change very often or ever, so it's not efficient to fetch the project_id every time we create an issue. But that's what we're doing. We should stop doing this and set the project_id once instead.
In order to not need a migration, we should make it set the project_id when saving the issue tracker (to make future errbits more efficient), and also fetch the project_id live (to keep already configured errbits working).