Re 2) When you use World Space, all vertices are transformed on the CPU, thus the world matrix for that mesh essentially becomes the identity matrix. It sounds like what you expected, but didn’t get, so I’ll take a look and try to create an example for you.
Re 3) The batching mechanism triggered only when using World Space. When using “Local Space”, each draw call uses a separate (unique) world transform, and thus cannot be batched. For that to happen, we’d have to implement instancing, which we currently haven’t had in the road map.
I added #4818 for instancing, and #4819 for physics custom meshes using the buffer format.