DatastoreAccess on one cluster node is closed probably by ClusterSanityWatchdogInstance (in fact: SxpControllerInstance) when leader is isolated and new leader is elected.

Afterwards it is not possible to:
Delete SXP node which was created by DatastoreAccess on leader before isolation.
It is possible to delete that SXP node from another cluster node.
But it is possible to create new SXP nodes and delete them on the same cluster node.