Wilson-
Other than the obvious problems with the "no metal" rule

I do have one serious question. How will the roller axles be attached to the hub, and have you looked at the stresses across that joint? From your hub design, it appears that the rollers will be split in the middle, with half-rollers cantilevered out on either side of the tang through which the axle passes. As the wheel rotates, (assuming a proper profile on the rollers - which many of the mechanum wheels used by FRC teams do not have) the contact patch between the wheel and the ground should start at the end of the roller and move across the entire length and off the other end. When each roller first makes contact with the ground and the contact patch is at the end of the roller, the torque applied to the axle/hub joint can be significant (for a typical ~150 pound FRC robot, 600-1000 in-oz would not be unusual). That obviously is a dynamic load as the contact patch moves from one end of the roller to the other. So expect those values to at least double as shock loads if the roller is not shaped properly and you get intermittent contact between the roller and the ground as the wheel rotates.
Many designs avoid the center joint/cantilevered roller approach and go with an axle that is supported on both ends. This can add in a different set of issues (e.g. the axle supports must be designed to avoid striking the ground when the wheel rotates), but it does avoid one of the most common failure modes. The alternative is to really beef up the axle/hub joint to make sure it can handle the loads you are going to see.
I am not familiar with the mechanum wheel design you have used previously, so you may have already addressed this. In which case - "never mind." It is just not obvious from this image how the design handles this.
-dave