I don't know if this is just the way the Colorista II keyer works, or if this is a global After Effects-wide order-of-operations thing that I don't fully have a handle on.
As far as I understand the SDK, it is not considered practical, since it would require a plug-in to establish its own comp buffer outside the main rendering pipeline, which could lead to issues with color consistency and things like premultiplied vs. straight Alpha. I also seem to remember that as of CS5 it is actually no longer technically possible. Hence effects that use specific source footage API calls will bypass any previous effects and directly operate on the source.
My question is, what (if anything) is the difference in behavior between multiple effects on a single adjustment layer vs the same effects spread out over multiple adjustment layers? Is it predictable & consistent?
The result should be the same under all circumstances, assuming all adjustment layers have the same opacity and no masks and their stacking order is equal. Only when you apply masks or adjust the opacity to fade specific adjustments you will see differences. It is also to note that having all effects might be a bit faster and more efficient, since less data is passed from buffer to buffer...
I'm just saying, for certain, that the keyer in Colorista, when operating in the first adjustment layer (with or without other effects before it) will use the un-effected source to pull the key, whereas if it is on a second adjustment layer, it uses the source plus the effects on a lower adjustment layer to pull the key...which, in my case is a good thing, because I need to be able to use things like Levels & Hue/Saturation prior to doing the key - and doing that all on one adjustment layer doesn't really work.
As for it being faster & more efficient, I don't doubt that for a minute. AE can turn into a sluggish beast when doing a lot of color correction.