Thanks everyone!
JakubCech: I actually did not use any id masks or render elements. My fusion comp was mostly super simple. I have outputted only 1 render element! Zdepth pass :D
What's in my comp is mostly just some glows, glare, Dubcats magic photographic LUT, some Fog using Zdepth pass, and then also volumetric clouds in fusion were done just using Zdepth pass for depth masking and FBX-imported camera from 3ds Max scene to match the perspective. If I did some breakdown, it would show pretty much the same render, except you would see some color corrections, little bit of glow, atmospheric fog and clouds popping up :)
Melviso:
Yep, I did just quick and dirty job of rigging and posing the guy, so it doesn't look the best. Will try to improve it next time :) I guess what looks off is that his upper leg is quite high up in the air, implying he is still alive and has tensed muscles. If he was truly dead, the pose would be lot more relaxed with pelvis dropped all the way flat on the ground :) As for motion capture, I've toyed with it a little bit, but always smashed my face against some CAT bug :)
Ałtaj:
The base terrain was done in worldmachine. Then I selected slope polygons and extracted them into more dense meshes. Which were then displaced using procedural mixes of bercon noises, which were then decimated with pro-optimizer to keep the polycount reasonable. It was textured using complex procedural shaders, mostly utilizing Corona Triplanar map :)