For the same rider/bike weight and same speed the energy needed to remove to slow down should be the same, regardless of the brake in use. However how that brake design handles the heat will be different (from a coaster to a roller). How much thermal mass (how much metal and what kind), how quick can the heat be removed from the brake (cooling) and how the heat affects the other parts of the rear hub are what I see as the aspects that count (independent of how well the roller brakes works as a brake if heat wasn't an issue).
I will suggest a roller brake has more mass than that of a typical coaster brake. That the roller brake housing is a separate and removeable part of the total hub has me speculating that the heat won't transfer from the brake (to the gearing/bearings) as quickly as with a coaster type. I also suspect that with its larger housing and sitting outside of the spokes the roller brake will cool at a faster rate than a coaster. And given all that I suspect the gearing and bearing lube in the rest of the hub will see less heat and thus work as a lube better. One could use grease with a high temperature capacity (as Shimano specs) in the roller housing and a different lube in the rest of the hub (again this is how the Shimano roller units I have dealt with were done from the factory).
But I'll add more opinion
Roller brakes, as I know them and their history, were really intended for a more transportation oriented application. All weather, third world, urban commutes and the such. Where speeds tend to be lower that some MtB trails can see. Where the distance of applying the brakes is relatively short (parts of a block VS down a 2000' decent be that on or off road). So both the brake's horse power and its heat capacity don't need to be high. Additionally the design of having a roller (or 5) get wedged between a moving and a stationary set of ramps means there's little "bottoming out" of the ramp attached to the cable. Think about the angle the ramps are set at to each other and the flat contact other brake designs have. I could do the trig for the loss of contact pressure VS cable travel but will leave that for others. It is this aspect that I dislike. The initial lever/contact point will see little power compared to disk or rim systems (remember in my world the coaster brake isn't on the list anymore because of the lack of free back pedaling).
But I have learned well how to not call other preferences wrong, safety excluded (hence my ft brake question). Andy