Consider the case where we have a master and slave Karaf instance:
1. Master has acquired lock
... time passes...
2. Network problem happens
3. Master loses lock (due to the network problem) and reduces start level accordingly
4. Before start level reduction is complete master manages to grab the lock
6. Start level is set back to normal
Seems that flipping the start level around before the container has time to complete is not safe. Proposed solution is to just pause the locking loop until the start level is actually complete before trying to regain the lock. This will prevent flipping of the start level too quickly. Patch coming soon