Use of solid brushes and CSG in level building

Since Quake 1 (I think it was the first), level design tools for FPS games have used solid brushes and CSG operations to define level geometry instead of more conventional poly modeling tools you see in general purpose 3d applications like 3DS MAX.

Why is this? Is it:

• More productive for the level designer?
• More likely to lead to closed geometry that won't screw up potentially visible set pre-calc?
• Results in lower poly count (more cumbersome tools dissuade designers from adding too much detail)?
• ...

From game development U62

• Mostly, in practice, it comes down to optimising lighting and visibility calculations. Portalling up a CSG world is relatively easy compared to polygon soup.

  It doesn't usually end up with a much lower poly count because many people who do use it allow decorative objects to spice up the world after building the body of the level. It's not normally better or worse for a level designer, but does allow a more "rough it out" approach than traditional mesh builder level design.

  U62 : I watched the 3DBuzz tutorial videos that came with Unreal 3 and they mentioned that designers usually cover the brush geometry with imported meshes. So it's almost like the system exists to provide 2 layers of geometry - a rough layer that's used for visibility calcs and a detailed layer consisting of imported models that are occluded by the world's brush geometry, but don't occlude anything themselves.
  From Richard Fabian