Recovering from a Single-Node downtime in a Multi-Master Setup