At first glance, this looks like the issue may be with the available space in the /tmp/var/target… directory. Can you check if there are any space restrictions you are running into on that machine?
You can also try to manually scp the files, for example libcuda-injection.so, into that directory and the result may shine some light on what’s happening. In theory, you should be able to manually copy the entire target directory over and then when you run the remote profile, it will see the files are already there and not need to copy anything at all. If you do copy everything, the ncu binary will be there and you can try logging in and running a CLI profile directly on the remote machine.
If space is the issue, you can change the deployment directory in the GUI when setting up the remote connection to something with more space, like your home directory.
Let me know if any of that helps or if you still are running into an issue.
I encountered the same problem. Even after manually moving the files, the issue persisted. I can confirm that the space size and permissions are all correct.