The workaround is to repeatedly do "bzr break-lock" until it no longer prompts the user to break a lock. At that point you can assume any backlog of stale "bzr serve" processes that are still waiting on the lock will have all acquired the lock (and had their locks broken).
The steps to reproduce work.
The workaround is to repeatedly do "bzr break-lock" until it no longer prompts the user to break a lock. At that point you can assume any backlog of stale "bzr serve" processes that are still waiting on the lock will have all acquired the lock (and had their locks broken).