i would have to imagine that we want to cell to be " out of phase" aka apposing voltage potential....
At a glancing look I can see this to be true, but when you think about it a little deeper, we really want the two chokes in-phase. Remember though, they can never be exactly in-phase because L2 has fewer turns. So L2 will effectively finish its cycle time before that of L1 and the secondary. If L2 finishes first, then it is already into the next cycle before L1 and secondary finishes. This is the misalignment I'm talking about and I think this is where the polarization comes from, because what I see happening is each cycle actually resets by the VIC driver. The VIC driver is synced to the feedback of the core which comes from the half of the core with the primary, the feedback coil and L2. So what really happens is L2 sets the cycle time and L1 and the secondary never get a chance to completely finish a complete cycle. So when L2 gets back to the zero volt mark, L1 and the secondary are still lagging behind with negative volts. We can flip that by starting the signal low, then ending high--basically just move in the wave by 180 degrees. In this case when L2 completes a cycle and is back at zero volts, L1 and the secondary are lagging behind with plus volts. So to get optimal voltage from each cycle we would want L1 and the secondary lagging behind by 90 degrees, that would put it right at the peak voltage when the cycle is reset.
I think it is important to notice why L1 and the secondary are on one core half and L2 and the primary are on the other. If they where on the same core with no gap, the VIC driver would attempt to force them all to follow the same cycle time, but with the loose coupling of the separate core halves and the gap, L1 and the secondary tend to oscillate at their own natural frequency which is lower than L2 and the driving signal of the primary.
I also think the diode between the secondary and L1 reduces much of the mutual inductance so that each coil tends to operate independently of the other, each at their own natural frequency. Otherwise those two coils acting together as one big coil would have a frequency way lower than L2, probably less than half. Ideally, each coil should oscillate at 75% of L2 to get maximum voltage polarization for each cycle of the VIC driver.