I have the same issue, but in my case I deleted the VM from the vCenter in an attempt to get the NSX manager to fail and let me retry. It never did. I tried multiple reboots and have even gone into the CLI on the managers and confirmed that the node is removed.
No luck. The database seems to be keeping a record of the installation status and the option to terminate and delete is not available.
I was able to deploy a 3rd controller using a different name and ip as the in use name and ip will not release from the stateful database.
I am doing this in the NSX-T 2.4 beta lab and my resort to a clear installation of everything to see if the issue can be resolved.