On the VNM, in the SpectroSERVER configuration, there is an option to use loopback when modeling. If this option is enabled and a discovery is run using an IP range, a duplicate model will be created if that range includes the management IP of that device.
When configuring VRF, the management address is not added to the IpAdEntAddr table of the device causing Spectrum to believe this is a different device.
This is resolved in 23.3.12 and above. The following needs to be added to the $SPECROOT/SS/.vnmrc:
preventDupDisc=TRUE