I am going through the pinboard tutorial and there is a section where he is trying to demonstrate that you can improve the effect of the flat lighting being rendered by attaching an xform and a point op and then translating the normals.

But the normals on one of the grid surfaces are now pointing away from the camera, that doesn't make sense does it?

I am not too familiar with that specific technique of creating the pins but normals pointing away from the camera will certainly shade the grid darker where you would expect it to be lighter and visa versa.

Sometimes normals are hyjacked to do some specialized task. Normals are handy to use because you can visualize them in the viewport directly. After you use the normals to the work, you can always put a point SOP at the end to either remove the point normals or rebuild them with a facet SOP.

With the layer SOP you can add any number of Normal layers.