Using Raid 0 in Windows is not a good idea. If one of the vmdks fails the data is lost.
So you would rather use Raid 1.
But then you get the new problem that a VM will not boot if one of the configured vmdks is not present.
There is a workaround for this: if you configure your VMDKs to be USB-disks the VM will still boot when one of the two is missing.
But I doubt that you would really want to use vmdks configured as USB in production - just for performance reasons.
The problem with the one missing vmx-file is quite trivial.
You could use a script that checks if vmx-file 1 is present: if it is launch it.
If it is not present, check presence of vmx-file 2 on the other datastore. If it is - launch it - if not throw an error message.
The idea is nice - but honestly I never heard that any self respecting admin goes that way.