The question Cactus posed was: why ever use a stereo track? I can think of only two reasons: either you're dealing with a prerecorded stereo sample, or you want to be sure to maintain the precise relationship between the two microphones as recorded, throughout the process.
As for why SSL chose the -4.5db center, I'm guessing it's just a compromise between the two most-common pan laws, -3db and -6db. On hardware consoles, the pan law is usually fixed, determined by resistors soldered in place. One group of people prefers -3db center, another group likes -6db, so the manufacturer says here ya go: -4.5db. That'd be my guess, anyway.
IMO the "correct" pan law is either -3db center, or no correction at all. The former emulates how sound really moves through real space, the latter accepts that it's not a real space at all and therefore real-world rules need not be applied.