As is the case with any sufficiently complex thing, it takes time for all of this to become internalized / second nature.
In your situation, the base object (the deformed cube) was not just scaled to become elongated, but at some point you also translated one of the faces giving the base object the sheered look. This is one reason to use multiple views of your object to help catch deformations which might be tricky to see from just a perspective view. (Over time you will see all of this immediately.. but it takes practice)
Once you correct the sheer (best just to use a new object, or learn how to snap edges to the (x,y,z) coordinates of other vertices) , you'll still get deformation in those parts of the track which bend a lot. This might be OK if you're aiming for non-photo-realism, else there are hefty tutorials for tank tracks.