Well, uncovered a couple of things. There's not much of a curve to these rollers. Both are not flat, but the curve is very minimal, say 0.5mm. And apparently this set up wasn't adjusted to track well, as the belt sort of sat against the "L" and slowly ground it away. The belt ground through some sheet metal, and a some of the "L" behind it.
You can see the cut in the casting. It actually cut two screw heads as well as making a semi circle cut.
Here's the platen wear. It's uneven and varies between 5-20 thousandths.
Unfortunately, I think I have to mill this flat(ter). With the 7/32" platen, the belt won't be in a good place. When I put everything in place, along with the belt guard and the drive wheel, the geometry didn't look good. I am going to try to take off about 0.100" off the surface. That way the platen will only stick up 0.118" rather than 0.218. Probably make a pocket for the glass, with little cutouts for the corners. The pocket will act as a stop for the glass, should it come loose.
Since nothing on this piece is really flat, or for that matter smooth, I'm not going to bother getting 246 blocks. Have two 2x3x6 blocks of 6061 that can be used as spacers instead. The blocks were fly cut together to 'match' them and make them smooth and flat. I'll use the blocks to hold up the "L". Somehow I'll clamp the whole mess down and reference it to the table.
Here are the roller wheels.
The top wheel has a rocking adjustment, to sort of track the belt. It also has a spring to maintain tension on the belt. The bottom drive wheel is 5" OD with a 16 mm keyed shaft. The top wheel has a bearing, but I haven't looked at it yet. Have to say, grinding dust is terribly messy, and just gets everywhere.