Just a quick question, out of curiousity…
If I specify the same texture coordinate pointer for 2 or more texture units, does the driver recognise that it only needs to upload the data once to the graphics card (and share it once its resident on the card)?
I believe you only slurp it across the bus once, but the data probably gets replicated across resources on the card.
In the new ARB_fragment_program world, you actually have access to separate (orthogonal) texture coordinate sets vs texture image samplers, so it’ll be less of a problem there.