Agreed: The on-camera system probably sends back one power level for each group. This still leaves unanswered the question: what if each remote group has multiple units?
On several of my law enforcement night shoots, I would have two or three distinct, separate groups: a distant patrol car with two or more lights on it, a backup officer with one or two lights on him, an arresting officer at the window of a stopped car with three or more lights. In these setups, the light from any one remote group would have little, if any, effect on the other two groups due to distance and setup. In all my night shots, the remote SB800s are off-camera, so the system cannot identify the location of any lights. All the camera can see is the light reflecting off the subject(s). How does the system know that there are three distinct subjects, instead of just one? How does it know if the light from all three groups is converging or diverging? Good questions.
In my top image at the start of this thread (all three groups in TTL), there appear to be three pulses emitted by the on-camera master for the three remote groups, and two pulses returned from the remote units. Look at the shape and intensity of the on-camera and remote pulses. Is it possible the remote pulses may include encoded data based on their perception of light reflected from "their" subject? Nothing to date suggests the on-board Camera/SB900/800/SU800 configuration cannot receive such data. The decoding logic is there when on-board SB900/800 units are used as master, and perhaps with the SU800 as well, even though it is never used as a remote unit.
The popup speedlights may be another case, which might account for the different algorithm you mentioned. Unless Nikon has isolated the bus from the PC port when in remote mmode, it may be a simple chore for Max to plug into a remote SB800/900 and take a look at the preflash pulses returned to the on-board master.
I may have to set up the above test again, this time with two or more remote SB800s in each of the three groups. This would tell us if all remotes within a given group respond simultaneously or not to the preflash sequence pulses from the on-board master.
If the SB600 ignores the 1/128 power level direction, and fires at the closest thing to it (1/64), it will be on stop too bright, which may be significant.
The deeper we get into this, the more questions arise.
HBB in Phoenix, Arizona Nikonian Team Member
Go here for a list of membership upgrade benefits.
Photography is a journey with no conceivable destination.