Even FLI can't make a Kodak/Trusense sensor noiseless, Greg
Dave is imaging with a KAF-8300.
I did an experiment with some narrowband data from my KAF-16803 at -30C and processed with and without darks. The integration using darks showed several percent better SNR. I get a few more percent from overscan calibration. Add a bit more for careful tweaking of rejection and scaling algorithms and now you're talking about an easily visible improvement.
Darks may be optional with the new Sony sensors. Mike S certainly seems to think so...
Cheers,
Rick.