When I’m rendering compositing work, I make sure to have nothing in the 3D space to speed up the rendering time but is possible to set Blender to render only the compositing nodes even if there is something in the 3D view that I don’t need in my comp?
The compositing is considered as post processing after the 3D rendering in Blender, can’t it be the main process?
“RenderLayers” are, generally speaking, a filter for choosing particular things to be rendered and for specifying particulars of how to render those things. What’s visible in the 3D viewport is an entirely separate and unrelated concern.
Yes, you certainly can set up a composite-node tree, using one or more RenderLayer nodes to provide inputs. But one reason why it’s most-often done in stages is to minimize (redundant …) computation time. In other words, once you have rendered (say) the background once, you might be able to use the same background in every frame thereafter if the camera doesn’t move. Likewise, you might decide to render the little bird and the ugly frog and the fisherman in three separate renders so that, if you decide to change the color of the fisherman’s tunic, you don’t to re-render anything but the fisherman (or, maybe, just the tunic).
It demands time and careful advance planning, but it saves hundreds of hours of computational work. (This is also what the “OpenGL preview” rendering capability is all about.)