Number one thing, to me, is to find if pulleys are all parallel and on the same center. (coplanar)
A laser line-level that you can lock in position (most of them) can cast an accurate line across the pulley centers from multiple locations. In FACT, a couple or more of them, preferably in different colors, simultaneously trained on multiple pulleys can help see if they are all on the same plane and if that plane is perpendicular to the axes of the various pulleys.
If, in order to get your laser lined up on three pulleys from one angle in dead center of pulleys results in a different plane than a laser lined up to hit dead center on three pulleys from a different location, at least one of the pulleys is cocked. The difference may be minute.
Another method is to use a straight edge to see the distance from pulley faces to center of each pulley is consistent.
If any of the grooves on any pulley feel asymmetrical, as in, kinda burr-y on one side of the groove only, it indicates a belt approaching that pulley from an odd angle, which should not be the case.
If this is a long-term problem, my guess is that the Kenne-Bell pulley is not lined up coplanar with the other pulleys, but just slightly, so the wear takes a while to show up. This is no slight on Kenne-Bell, disclaimer disclaimer. It is just the only part of the equation that is not factory, to my knowledge. Also, if you are using one of those slip-on pulley diameter-increasing pulleys, it could be a bit crooked. I doubt the tensioner is able to do this, unless it is grooved, also.
