My understanding is that the descent is initially calculated on actual wind at cruise level which decreases linearly at about 1-2 knots per thousand feet on descent. This is how the descent point will be calculated with no forcast winds entered on the descent forcast page. As the "beast" descends it will update the profile (VNAV path) based on actual winds. If there is a large variation in the actual wind compared to the estimated linear wind gradient then the "speed" at which the path descent was originally calculated will be sacrificed as I am sure you know. With the -800's, especially a heavy one, I have adopted the technique of entering a "false" tailwind component in the descent forcast page to get the aircraft on its way down a little sooner and at least keep the VNAV path without going into an overspeed disconnect. There will be better ways but I am pretty new on the machine and haven't had the benefit of anyones experience to learn from!
Incidently if you are flying an -800 into Sydney you can forget all this wizardry. ATC have a good look at your profile and will always chop off about 5-10 track miles when you are only 30 or miles from the airport. In an -800 at near max landing weight this becomes a real handful. Better to decline the vector I am learning.