The "ah ha!" moment for me was when I noticed that the clock position of the linkage arms to the bell-crank were very specific. A lever attached at 0 degrees (12 o'clock) moves much faster (back toward the firewall) than one attached at 45 degrees (but not twice as much) when the crank rotates, and of course one at 90 degrees doesn't move at all, and one at 91 degrees (past 3 o'clock) moves backward. As the crank rotates, the relative speeds of the attachment points changes again, but you need trigonometry to establish exactly how many thousandths per degree for both points.
