The reason is at some point you need to work out what your displacements are doing. Compute Terrain is there as a snapshot of the surface at that point in your work. It sets up the texture and normal mapping at that stage, for further displacements and colouring to occur. If you want one over the other, you use a Compute Normal or Texture Coordinates from XYZ. Compute Terrain encapsulates both processes.
Get Position, Get Normal have their uses, but remember they look at the final surface at render time, which when used for displacements often cause unwanted 'spikes'.
TG has always had the philosophy that you set up your base terrain, give it a general mapping via the Compute Terrain for you to colour and displace at a lower scale. There's nothing stopping you using several Compute Terrains/Normals in your scene, they just add more to the render time. There's no way of avoiding that.