I believe that it really depends on your perspective, or in this case the vCenter Server's. I would suggest that a file is orphaned (at least from the perception of a given vCenter) when it (the vCenter) is not aware of what purpose the VMDK file serves; or, to which virtual machine it belongs. This does not always hold true, but that's my general explanation.
Have you ever tried working with vCenter Site Recovery Manager’s built-in alarms? How about recovery point objective (RPO) violations? I have, and I quickly grew frustrated by the number of e-mails - essentially, false alarms - that I was receiving. When an RPO was exceeded - even by one minute - an alert would be triggered. While the alarms were technically valid, when you are protecting multiple VMs, and brief RPO violations occur frequently in SRM … well, you get the picture. Overwhelmed by these seemingly non-important alerts - "the boy who cried wolf" - they quickly became ignored. I needed a way to adjust, and fine-tune the RPO violation alerts so that I wouldn't miss something important.
