Actually, if you asked me, I'd say that the best way to work with DBE is by working each channel separately, for a bunch of reasons I've said several times in this forum already.
First, each channel is going to have different gradients, even when using an OSC camera, and each may respond better to different DBE applications (samples placed at different places, etc).
While DBE applied to a RGB image processes each channel separately, it builds its background model responding to the very same set of samples and parameters.
Sometimes this may not matter much, depending on your goals and your data, but other times it may. Foe me it matters every time.
Also, I like to "read" the background model, as it often helps me "guess" whether I'm removing JUST the gradient or something else/less, then go back to rework the samples and/or parameters if I don't like what I see. RGB composite background models often look rather psychedelic and are very hard to "read".
Someone may say that this could be solved by breaking the RGB background model into its RGB components, but then, when you go back to make adjustments, you're fighting three different channels all at once. Personally, that sounds like a much more difficult task than just applying DBE to each channel separately!!
Now, when you apply DBE separately on each channel - and that's how I do it all the time - the workflow seems quite simple and intuitive:
A slight crop on each master comes first, then DBE on each channel (where I spend as much time as necessary), then star alignment, some more cropping, RGB integration, and then finally BN followed by CC.
My opinion. Skin your goats as you like
