Just an update: I was able to delete the VMDK file of the failed conversion by going to the Datastore Browser. From the vSphere client, I clicked Configuration -> right-click the lone datastore -> Browse datastore and from there delete the corresponding vmdk file. The free space is back to a little over 1TB now. This is such a simple task likely known by most.
My concern now is whether to choose thin or thick provisioning. Does thin provisioning cause fragmentation of the vmdk file? I can easily choose thick as I expect no more than 4 VMs on the host, each one using around 250GB out of the total 1.2TB datastore on a HP DL380 G7 server. But I would love to hear your expert opinions first.