Well I've been really into the double camera stuff, and here's the simple answer:
The way we did it, each camera is completely independent, and works without knowing that the other camera is there. They both will convert both lights into a single blob. It would be up to you to write any calculations to find the location of the light.
http://kevin.org/frc/CMUcam2_demo.wmv Shows how the camera deals with multiple lights. The centroid is based on the number of pixels and the the concentration of pixels. So for example if you see two lights, one at an extreme angle and the other dead on, there will be a blob encasing both lights, but the confidence will be low and the centroid will be closer to the high concentration of pixels.
If you want to differentiate the two lights you'll either need to use the Virtual Window function (see the
CMU Commands manual), Or get a full pixelmap (not recommended).