Hey Andrew,
After reading the procedural generation suggestion another question came up. The drawback of the first approach (designing the level in advance) is that static meshes which are partially visibile (will be rendered as a whole and the parts which are not visible will be culled out), right? I would expect the engine to completely disregard meshes which are not visible at all at a given point in time.
Do you think that the first approach is going to be a performance issue if the game is targeted for mobile devices and not only desktops?