As much as we would like to think that CLS does compensate for multiple, possibly overlapping remote groups, I am convinced it does not occur.
Consider: On-camera master contributing to the image capture illumination plus three remote groups, A, B and C.
1) CLS does not know how many speedlights are included in each remote group. The on-camera master is obviously just one.
2) CLS cannot know if all speedlights are illuminating the same subject or different subjects.
3) The system knows how many remote groups have been selected prior to shutter opening. That information is sitting there in the on-camera SB900/800 or SU800.
4) It is not clear if CLS know the type of remotes responding to preflash pulses. They could be SB900, SB800, SB600, SB-R200 units. I suspect the answer is no.
5) The exact content of the apparent multiple pulses by the master and each remote group remains a mystery. From the all-TTL image (top) I captured, it appears that the on-camera master emits three brief pulses and the remotes emit two. We know that the on-camera pulses must contain channel number, group code and exposure compensation value at a minimum.
6) With reflected light metering, CLS has no idea of the reflective nature of the one or possibly multiple subjects. Hence, it tries to make everything a neutral gray, group by group, based on what is reflected.
7) Trying to guess at overlapping illumination and subject reflectivity would likely lead to chaos. Given the above, I suspect that metering each remote group independently and leaving exposure compensation adjustments up to the photographer is a wise decision. After all, it is quite simple to do this.
8) Until I get my hands on a four-channel recording oscilloscope (not soon, given ongoing NAS priorities) this is as far as I can go.
9) I suspect the reason for delivering the exposure compensation settings prior to shutter opening is to avoid adding unacceptable delay during shutter open intervals, particularly at the max sync speeds (e.g., 1/250, 1/500). Assuming my rotational velocity is relatively constant and scaling the above images, the preflash sequence occupies approximately twice as much time as the shutter open interval of 1/30 sec, making it 1/15 second. Assuming that it takes one-half of this time to transmit the exposure compensation data to the remotes, this would be about 1/30 sec (0.033 sec = 33 milliseconds). When shooting at 1/250 sec (0.004 sec = 4 milliseconds) with on-camera master and three remote groups firing, there clearly is not enough time (4 milliseconds) to deliver 33 milliseconds of data.
10) While circuit times are very fast, they are not instantaneous. While the remote speedlights appear to be firing at relatively low power during the preflash sequence, they may need a bit of time to recycle in case they will be firing at full power during the shutter open, image capture interval.
11) I presented the above two images at 16 by 20 inches on easels at Winston Hall's Nikonian speedlight workshop here in the Phoenix area last weekend and we had a great time discussing them.
Thanks to all for your comments and discussion.
HBB in Phoenix, Arizona Nikonian Team Member
Go here for a list of membership upgrade benefits.
Photography is a journey with no conceivable destination.