do I have to be concerned about csg rendering more than one face for performance
Generally it’s fine unless you use very complex meshes or a huge amount of simple ones. One thing to note is that CSG is mostly meant for quick level prototyping, but once you are done with that you can convert the result to an ArrayMesh to save on calculations and improve the performance.
You can then even export the resulting mesh to glTF/ glb and edit it in Blender etc. for a polish pash, which is a pretty nice workflow since it gives you both the quick iteration time of building your levels directly in the engine as well as the extensive set of tools that comes with Blender or other modelling tools.