I took the Comber to Cabin Fever, as Elmer's Engines are a theme of the show. I ran it periodically until it threw a wheel (cotter pin on roller axle came loose). One of the exhibitors at a nearby table had a double size Comber that ran very well at slow speed. I noticed that his forks didn't extend past the edge of the cam, hence less friction. Seemed like a good idea to try, as it also makes assembly easier.
I also decided to do as Gail and use ball bearings as the rollers. Way back in the early days I'd bought a package of such bearings in anticipation of the Kerzel team build that never got underway, and even better managed to find it. The bearings are .25" in diameter, so equal to the roller. Shaft diameter is 1/8" and width is .100": perfect.
I made the first test fork and installed it, at which point I realized that the forks are needed to keep the rollers aligned to the cam. Not sure how the other guy did his, but at least one of the forks needs to extend past the cam edge. So for the second fork I took the old one, drilled the roller axle hole out to 1/8 and used some 1/8 drill rod as an axle. The drill rod is a press fit to the bearing, so no retainer is needed; the fork keeps the bearing centered in the slot and the axle secure. I actually milled down the inside arm of the fork enough so that the cylinders could be squeezed into the cam.
The engine now runs a lot better after firing it up. Previously needed 40 PSI and ran very fast. Now at 20 PSI it can run much slower, probably around 200 RPM.
I measured the valve diameter (.499) and the diameter of the bearing (.505), so the excess air consumption is pretty easy to understand. To improve this I could remake the bearing with a tighter tolerance, but for now I'm going to let it be. If I decide to go to NAMES by air, I can take it and the Coventry as carryon.