This makes me think about removing every other sprocket tooth and then milling down the attack angle on the forward side so that the two rollers would both seat

This makes me think about removing every other sprocket tooth and then milling down the attack angle on the forward side so that the two rollers would both seat on either side of one tooth. I think the point of this design is that the link transfers compressive force forward as pressure if both sides are engaged. The loss in a chain is nearly all in the roller as it engages with the first tooth in the front chain ring. Most of the force is being transferred from this first tooth, and it presses directly on the roller, which is also rolling into its seat position on the tooth. So a better design would have the roller get seated on the teeth before the force gets transferred to it, and then have no rolling until force transfer is done, then roll off gently. A new chain on a new chainring engages at the bottom of the tooth, right where it seats, and the torque of the chain ring rolls it up the forward tooth ramp until the tangential component of the force pressing into the contact point equals the pedal force. This happens on all points engaged on the sprocket. So you get a lot of rolling, but that’s over a lot of rollers. As the chain wears, it gets longer, so the teeth catch early from the tip, and it rolls all the way down the tooth while the entire chain tension is all on that first roller. This particular design has no rollers at all. It will still stretch because it has the same pin-plate interface, and it seems that a modest stretch may render it inoperable, but I’m not sure.