Wouldn't cloning the vmdk and storing it on a tertiary location work for you?
No, if the data is in the 'detachable' volume it is much easier to handle, we may detach the volume from a VM and attach in another running VM instance.
Data is backed up only at the time of cloning but if there is a separate volume, data stays there all the time.
So long as you only want the data to be accessible by a vSphere VM, you can do this using regular virtual disks - just build the original VM with at least 2 virtual disks, 1 for the OS/app and 1 for the data.
You would then have 2 sets of VMDK files, they would live in the same folder on the same datastore by default, you could put them on separate datastores if required.
If the time comes when you need the data disk to be accessible by another VM, you would add a new disk to that VM but using an existing VMDK file (rather than creating a new VMDK, which is the default) - you would be prompted at that point to browse your datastores, find and select the VMDK file from the original VM.
If you want the data to be accessible from outside any VM, you could use an RDM for the data instead - that’s where your virtual disk is a LUN rather than a VMDK file.