MoveReplica API for shared file systems attempts to delete the old replica and then creates a new replica (with same core and coreNode name) on the target node. If the overseer fails between the two operations then the replica is lost. The API should detect that if the source is not live, it only needs to create the new replica. Then the old replica (upon coming back online) auto-detects a replacement and unloads itself. This is also how OverseerAutoReplicaFailoverThread works today.