Ok, I'll try to make myself clearer...
A multipass picture is, as you know, several renders of the same "view" with different settings and combined through filters, with an external "combiner" like photoshop or premiere to do the "filtering" you can adjust each layer and have a much greater control of the final outcome...
Imagine already renders the "view" several times when it needs to do certain effects like "motion blur" and "depth of field", the difference is that the renders are always of the complete scene and the filter is always to just mix the results up without any "per render control".
But since it is already able to call upon its renderer and combining results, I believe it would be possible to use the same functionality to do "true" multipassing, maybe not with the final control of each layer as you would have with the separate images but that might be possible to.
If nothing else, it might be possible to let it automate the multirendering automatically and produce the different sources for the external software without you all needing to set it up, thus making multipass more available to the masses.
All this of course falls to the source, which I haven't seen yet.
If its still isn't clear, well, you can blame me, I might misunderstand all of it as usual