I really want to avoid getting into a back and forth internet pissing contest on this so I'll sign off with this and be done with this.
If you're familiar with mathematics, think about the geometry involved in terms of a Cartesian coordinate system.
In this system, a point in the plane can be described by 'x' and 'y' coordinates. A line from the origin of radius 'r' has its end point described by:
x = r cos (ø)
y = r sin (ø)
So picture the tie rod pivot point at the rack as the origin of this plane. The length from this origin to the ball joint on the spindle is 'r'. The terms x and y describe the position of the outer ball joint in the plane relative to the pivot point on the rack. The angle ø is the angle off horizontal of the tie rod.
If you were to disconnect the tie rod outer ball joint from the spindle and swing it up to its range of motion upper limit you would find that while 'r' does not change, x and y do. The x coordinate of the ball joint -- projected down to the x axis -- decreases as the angle ø increases: a plumb bob on the outer ball joint hanging to the ground would move inward toward the center of the car as the tie rod is swung upward.
The key thing to note though is that because the x coordinate is based on the cosine of the angle of the tie rod, the amount the x coordinate moves is not linear as the angle of the tie rod. For example, assume 'r' is 15-inches. For an angle ø of 0-degrees (horizontal), x sits at 15*cos(0) or 15". If the tie rod is moved upward to an angle of 10-degrees, x now moves inward: x = 15*cos(10) or 14.772". This is a difference of ~0.228" and is a decent approximation of how much the wheel will be toed in when the
suspension deflects enough to give a 10-degree tie rod angle. Given that toe angles are typically measured in 16ths or less, 4/16ths of an inch is actually fairly substantial.
But what if the tie rod is already at, say, 20-degrees due to a
suspension drop and that same 10-degree
suspension transit is performed. At the start, x with no additional
suspension drop is 15*cos(20) or 14.095". (NOTE: This, of course, would require a lengthening of the tie rod via adjustment to bring the static toe back into spec. This results in a larger 'r' but I'm ignoring that for the moment.) Apply 10-degrees more upward movement and x drops to 15*cos(30) or 12.99" This is a difference now of 1.105" or nearly 18/16", more than 4x more than the previous example. So the wheel is toed in over an inch during this
suspension movement on the lowered car versus just under 1/4" for the unlowered car. This is definitely going to feel like **** on the road.
So the same 10-degree absolute angular change brought by
suspension movement results in drastically more toe if the tie rod is already angled. This is, of course, because the derivative (rate of change) of x as you move in fixed ø increments around a circle
increases as you increase from 0 to 90-degrees. It's easy to conceptualize this or draw a circle and play with a protractor and a ruler...
Bringing the tie rod angle back to close to zero is critical for minimizing the cosine effects described above. That's what these kits are intended to do.