So, I want to render my terrain from the point of view of my light source. It seems pretty excessive, however, to do all the Vertex-Texture-Fetches associated with geoclipmapping twice.
That is what you want to do, and that is how it is done. The terrain is drawn from a different perspective in the shadow mapping pass (from the sun's perspective) and the texture fetches and results of the vertex transformations are different from when you draw the terrain from your camera's point of view.
However, you can render a 1/4 polygon terrain for the shadow map pass, and use a filtering technique that softens the shadow edges (VSM/ESM and variants of these), and the result will still be good.
I'd also have to set up a special vertex shader for the shadow-mapped terrain portion (to transform the clip windows appropriate to the camera and sample the textures for the height values). This would be different than my normal shadow vertex shader (which just does the basic gl_ModelViewProjectionMatrix * gl_Vertex for regular objects).
To draw your terrain from a new perspective, you just pass in a different world-view-projection matrix. You don't need a different vertex shader (you do need a different pixel shader, because all the shadow map pass pixel shader outputs is a depth value).
However, I would make a different vertex shader, because the only data you need from the shadow mapping terrain pass is per pixel depth values. So you can render your terrain without any textures applied, and no normal map lookups are needed, etc. Depending on your needs, you may choose to forego ring fixups or whatever you use to avoid cracks in the terrain.
Overall you should be able to get away with a much faster rendering of the terrain at the shadow map pass than your usual textured+lit terrain pass.