Now, we only synchronize jobs when there is a coordinator address update if node passes from master to slave, but also if a node is a slave of a master and it changes to be slave of a different node jobs should be also synchronized.
Steps to Reproduce:
- Start four nodes.
- Force them to lost visibility, so that two subclusters are generated: [node1, node2], [node3, node4]
- Make them be part of the same cluster.
If for example, new master is in [node1, node2] only subcluster master node in [node3, node4] gets synchronized against new master.
All nodes that are in subcluster different from the one that contained the new master should be synchronized, not only old master nodes.