Well that was unexpected

CS lightmap updates are slower than CPU lightnap updates.

I really need to think about that one before I say anything more.  Maybe the code I wrote is shit or maybe switching over to CS mode mid-frame is unhealthy or maybe it's something entirely different.

But it caught me by surprise all the same.  Just glad it happened early enough.

This doesn't mean that I'm completely giving up on this idea - just that the more direct route has been proven to be unviable.  There are other ways.