Paul and Doc,
The pull on the leadouts almost never goes to zero on one line. When it does, you've hit the 'Netzeband Wall'. (Where the pull required on the 'acting' line side of the bellcrank to turn the model as you want is more than the total pull the model is generating at that moment.) That is the only time you're actually 'hanging' the model on one line, and you can really see the unloaded line droop way aft from its own air drag. Yeah, I have done that once or twice... very disappointing...
Doc, I have a little more theory on the separation between the leadouts at the guides... With the lines UP forward/ DOWN rear, if you can estimate how much the model pulls before starting a, say, square corner, and presume that all of the pull does shift to that line for an instant during the turn, you can also estimate how far forward of the mid-line, at the line rake angle, it takes to neutralize the gyroscopic precession yaw for the RPM and prop. Precession can be estimated. So can line rake.
If my ideas on this are right, or even close, leadout guides places over and under do nothing to oppose, much less cancel, the precession forces. Further, if the model rolls in flight - from a gust, or from a rapid yaw making the wings change relative speed, that creates a control input through the bellcrank.
Clearer, (perhaps?) if you're flying level and start a sharp corner, there'll be a gyro precession yaw unless something prevents it. For an UP-control input, the nose swings right. The left wingtip move faster than it did before the gyro kick. Since the model is yawing on the CG as center, the right wingtip slows, compared to before the kick. (Yes, the speeds 'progress' across the entire span, but the tendency is the same.) When a wing moves faster, it lifts more; when it slows, it lefts less. That says the model tends to yaw and roll in response to a sharp input. (Can you say:"Hinging?")
Overn and under leadouts act like a handle input when the model rolls... With the leadouts horizontally placed, there is some of the same tendency when the model yaws, but we're used to that. It always did since we started flying CL...