Say you have your nodes set up, maybe a video and backdrop. The what do you do? Click ‘Animation’?
Press ‘Animation’ to render an animation or ‘Render’ to render a still
Think of it this way: “the point” of compositing is really, “rendering in stages.” You use the renderer(s) to produce components of the final scene, in a way that you have planned in advance. Then, you use compositing techniques to combine them together, and to alter them … much as is done in a darkroom.
Now, you can also apply these same ideas in one render network, which (say, through RenderLayers) “splits the job into several parallel data-flows and then, at the end, combines them all back together.” Sometimes you do things like that just to get around the pesky restrictions of physical reality … knowing that “CG doesn’t have to be real, it just has to come out plausibly-enough looking that way.”
Usually, you’re looking, not only for creative-control, but also to minimize the amount of computational work that you might have to (throw away(!) and …) re-do. So, in any project or shot that uses compositing, you’ll be pressing “Render” several times, saving individual sub-outputs to individual files or sub-directories, then, in a planned later stage or stages, bringing them together … the hope, at least, being that you won’t have to re-do something that’s already “in the can.”