Compositing Question

It depends, it just needs to be rendered once, if your render will be piped through the compositing nodes and what ever you have in the compositor will be burned into the render. I do not take that approach, I prefer to render raw and then take the images into the compositor and then apply the effects and re-render again, that way I have much more control and the render time is just a fraction of the actual render. There is no wrong way though, you can take whichever approach you like.

You will need to render once to view the image in the compositing workspace, then when you finish the compositing you will need to render the image again and save your finished work with compositing applied.

If you are using a Viewer Node (make sure it is connected the same as the Composite Output), you can simply Save the Image in the Image Editor, after you finished Compositing; no need to re-Render.

Really?
All these years I have been wasting my time rendering again!

Although, most of what I do is image sequence or mp.4 and only render one frame before compositing.

On my 'old' computer I often had Render times of 1 to 3 hours (for a still), so that was quite an important realization, when I found this 😉

I think I understand the nuance of your question EEric. I've been rendering sequences for our latest Collab project for the past month. The workflow has involved rendering every sequence twice - actually 3 times.

It depends on the complexity of your composition. For this course (and 95% of every project I do), it's simple enough to pipe my render output into the compositor and render+composite at the same time. I.e. one rendered sequence or even video file output if it's fast enough.

However the collab project is much more complex and I need to do funkier things in the compositor than usual, including passes and ID mattes, etc. So in these more complex situations I render twice (or even thrice):

1. Raw 3D render data including passes and layers outputted as multiplayer EXR's. These sequences take the most time to render by far.
2. Composition where everything is beautified or motion blur added via nodes, lens flares, etc. When sequences take hours to render it's far more efficient to separate the composition from the raw 3D render. The output of this is a "final" high quality image sequence.
3. Comp image seq to movie file. This isn't always necessary but for this project I've been saving each sequence as a movie file for editing rather than editing with the image sequence directly.

If your workflow included an external compositor like Nuke, you would do the same thing of rendering .EXR's out of Blender and compositing with those sequences in Nuke.

Does this answer you question?