Quote: "1. Create individual objects at 0,0,0 and position them after creation
2. Create individual objects and set the vertices to the exact position I need them for the track.
3. Create an object and then keep adding meshes to the single object.
4. Create the track from, for example, objects of 10 meshes each."
1. and 2. will be equivalent, AppGameKit doesn't mind if the vertices are far away from the object center and the mesh bounding box for frustum culling is close fitting.
3. will be faster up to a point but I'm not sure by how much. Meshes in the same object share the object and shader setup but adding a mesh results in rebuilding the collision data for the entire object. However if we're talking 100 polygons then that's nothing compared to rebuilding a 10,000 polygon object.
4. is probably a good compromise between the two, if you get into the thousands of polygons per object then split it into several objects to mitigate the collision rebuilding problem.
Probably the biggest impact on performance will be how many vertices you have in each mesh. If you have a separate mesh for every plane then it will result in a lot of draw calls which will hurt performance, although on desktop platforms you could probably get away with thousands of draw calls.
I haven't benchmarked any of this, so these numbers are off the top of my head, but as a starting point I would say keep objects below 1000 polygons in a single mesh, or as few meshes as possible.
