To reply to your questions:
A1: Metahumans can be customized with different clothing, so having separated skinned mesh is an easy way to swap them based on what you choose in MetaHuman Creator.
Not sure if they have a unique mesh with a unique skinning, and then simply transfer the skin weights based on your selection, but the skinning is consistent between different pieces of clothing.
A2: The head has a barebone body skeleton, that’s because the head follows what the body is doing, and the joint of the face are driven by the actual facial animation.
If you take a look at the AnimBP of a Metahuman face, you’ll notice that there is a Copy Pose from Mesh node, which takes the animation data from the body and applies that to the head, then you have a Layered blend per bone, which pretty much separate the incoming data, so that the body animation is applied up to those joints ( neck, facial root ), and everything above that is driven by the face Control Rig or your standard animation sequence.
Also the Face is a child of the body, and because of the Master Pose setup in Blueprint, it will follow the body, along with the body animation from the AnimBP setup.
Issue2: Best way to have a fully rigged face based on Metahuman’s skeletal structure is to use the Mesh to Metahuman tool in Unreal engine, there are plenty of tutorials about it.
Then you can download the source file for Maya, so that you have the rig ( anc Control Rig ) in Maya as well.
QuestionB: Based on LODs ( both in Unreal and Maya ), corrective joints from the face are removed, so the face skinning is simpler.
Only LOD0 has all the joints, LOD1 and so on have less and less joints influencing the face.