Remote development using a firewalled target

I am using Nvidia Nsight 10.2. I would like to debug C++/CUDA code running on a cluster node firewalled behind the login node. My reference read is the blog but in that case the target is pingable.

With Paraview I can work around this because it enables to set up a reverse connection: the target cluster node can see the host node and I start the server remotely and then hook the client/host to the server/target. In that way I can leverage the GPU on the remote, with the appropriate extension plugged in ParaView.

Is any possibility to work around firewalling in Nvidia Nsight when it comes to code development, or there are elementary means to make it work that way? Thanks for considering this.