The server should re-send change notification for entities that the owning resource failed to replay. This is a more generic approach to re-creating items without RID.

The notification bus is bi-directional nowadays, so the resource could send a message when it fails to replay a change back to the server. The server would store it in a DB and would re-submit the notification later.

A few problems to figure out first:

1. How to store the notification on the server
2. Only re-send the notification to the owning resource, not other clients
3. When should Akonadi try to re-send the notification?
   • on next start
   • when the owning resource goes online
   • after a timeout
4. Initially re-generate the ADD notification for existing Items without RID