In rare cases, the vSphere inventory of managed virtual disks can become temporarily out of synch with datastore disk backing metadata. This problem can be due to a transient condition, such as an I/O error, or it can happen if a datastore is briefly inaccessible. This problem has been observed only under stress testing.
A discrepancy in the inventory can occur any time that a datastore is lost or when a managed virtual disk is created or deleted using a tool other than the vSphere Web Client or the VStorageObjectManager interface in the Web Services API.