borg operations always use an exclusive lock on the repository.

this goes back to a change when a deadlock issue related to lock-upgrade (from shared to exclusive lock) was resolved by just always doing exclusive locks.

always-exclusive locks can be a problem if one has long-running (or frequent) backup jobs (or in general: repo-writing operations) and in parallel one wants to do other borg operations:

• "read-only" operations, like borg mount / list / info / extract
• write operations (this is definitely the more critical / complicated case and could be omitted to simplify)

topics to research in this context could be:

• analyze / document repo locking requirements on a rather detailed level
• analyze / document other (cache? clientside indexes?) locking requirements on a rather detailed level
• can we have a deadlock-free locking without using always-exclusive locking?