Thus, we waste tons of animations to make special moves, like walking while watching by turning the head left and right.
Only two animations should be needed there, the legs execute the walking animation and the upper body execute the watching animation.
You are probably just doing it wrong since you really only need 1 animation and an aim offset.
Look into Unreal 4 animation layers. Google “ue4 layered blend per bone”
Don’t know about OP, personally I will use Aim offset and custom additive animation on top, since Aim offset alone may not be enough for lifelike animations. So it could be perfectly OK to have 4 animations for head movements alone. notably: look back right, look back left, (these two are difficult with just aim offset You have complex neck shoulder movements happening) look left - look right. If it really needs for extreme poses look up and down as well.
If you are making lifelike things, you would/should probably use morphs to fix the neck based on twist. It’s less expensive then blending more animations in additive…
Morphs are there regardless, not enough alone to correct animations, I was speaking of bone rotations themselves. But all this depends on how complex the character model is and how much time needs to be spent on it.