What's implemented in the VTP right now is a hardware application of the detail texture, as a second pass, with no masking, only one ground texture at a time. To allow the detail texture to co-exist smoothly with a lower-resolution geospecific texture (which contains the lighting, and ideally real-world imagery), the detail texture is blended based on distance from the viewpoint.
With the SMTerrain algorithm, distance to each vertex is already calculated by the CLOD algorithm, so that feeds directly into the alpha value of the detail texture.
With SRTerrain algorithm, we do not have a distance value, so the detail texture is drawn with a constant alpha value.
LoadNormalTexture(...)
Render(...)
LoadDetailTexture(...)
// turn on alpha blending
glPolygonOffset(-1.0f, -1.0f);
glEnable(GL_POLYGON_OFFSET_FILL);
Render(...) // uses alpha value per vertex, based on distance
glDisable(GL_POLYGON_OFFSET_FILL);