Hi, I am not able to answer all of your questions but I do have answers for some of them (I might be wrong about some info aswell)
So when exporting multipass in HDR (using ACEScg capture gamut) it seems that the final image pass produces the results that are not the same as in engine (it looks like your scene with r.TonemapperFilm=0 (it was default tonemapper before 4.15) but it does indeed capture with over 0-1) in After Effects I applied a tonemapper effect in order to replicate the results that would look simmilar to r.TonemapperFilm=1 (default in 4.15).
About anti-alias sampling and motion vectors. I personally turn the anti-aliasing completely off inside of the engine, I don’t like FXAA, TAA has too many ghosting artifacts in my opinion and just looks messy. So I guess rendering it at a higher resolution and downscaling it in Nuke would be the best choice (or even using some 2D anti-alias solutions in Nuke, I guess there is something like that). Oh and I don’t think you can export motion vectors (yet), it was recommended on the old UE roadmap so I guess it will be implemented in the future.
Something I didn’t find out is how to make custom render passes (something like World Position Pass), I used a post-process material that outputs World Position Pass and rendered it in HDR but I think it does not have all the color data necessery for making something like a position mask. If you know how to make a custom render pass I think you could make something like a custom pass for motion vectors (you can visualize them with this command: VisualizeTexture GBufferVelocity RGB UV0).
The multichannel exr isn’t supported (yet), but as for the motion vectors, I remember seeing this on old UE roadmap so this might be implemented in the future aswell.
Depth pass is clamped 0-1, and if using a atmospheric fog as a sky, the rendered image will have the sky as full black (tho this shouldn’t be a problem because there is basically nothing that has a 0 value when rendered, the closest what I could get is 0.0124 (and thats a gun attached to the camera). Even at 0-1 it still has enough information for you to work with it (I guess). U will just have to key out the full black and use something like levels to control the black and white point.
Heres the image of the stock rendered depth pass and an after image that has full black keyed out, is inverted and has levels applied to it.
Hopefully I answered some of your questions.