The frequency re-use is a factor of steering angle. Take these two spot beams, and the cells that they cover. The left side is a beam almost at nadir, the right side is close to maximum steering angle. Any hexagonal cell inside the beam’s footprint cannot be covered by another beam (from same or other satellite), as the overlap would violate the restrictions on Starlink’s license:
I don't disagree with the description of this type of spatial reuse. It will be interesting to see how nonlinear the capacity growth due to tighter angles is, however, and whether it's more like a 4x increase or an 8x increase.
The question of whether they can achieve their goals without additional reuse (for example, covering a given hex with more than one channel) is open. Fun times.