There are a few thoughts that come to my mind after watching these tests.
The cannon ball test may be a bit extreme, but it does provide a reasonable approximation of impact damage.
The failure is on the bottom side of the test panel, which is the tension side. The outer surface of the canoe, which is the side that is impacted, is in compression while the inner side is in tension.
For the most part, the wood makes very little difference in the results. The bond between the wood and the cover material may matter, but the penetrating epoxies and the porcupine roller seem to increase the strength of the outer fibers of the wood strips. That the porcupine roller would make a significant difference is interesting, since the wood is porous to begin with.
Since the failure appears to be mainly in the tension side, additional reinforcement on the inside may be more effective than reinforcement on both sides and should definitely be more effective than additional reinforcement on the outside.
Curvature does add strength, but is only significant if the impact is at a point where there is significant curvature. If you hit a rock with the bow of the canoe, there is considerable stiffness provided by the curvature. If you hit a rock amidships, on the side or bottom, curvature is not as much a factor.
The tests were conducted with the sample panel restrained on both edges. A canoe is restrained only by the water pressure on each side and the mass contained in the canoe. If the test panels had been supported by foam on each edge, it might be more representative of real world canoeing. A canoe can bounce off a rock. The fact that the cannon ball bounced several times on one of the test panels indicated that that panel was very rigid. Drop a cannon ball on to a piece of foam and see how many times the cannon ball bounces. The foam won't be damaged because it absorbs the energy of the impact. This may explain why the Royalex sample appeared to perform well. The skin is resilient and the foam core allows considerable deflection. Fiberglass epoxy systems are fairly rigid and won't allow for much deflection before failure. Kevlar should handle much more deflection before failure and should make a significant difference when used on the inside, the tension side. The epoxy is the binder, while the fabric provides the strength. I think that Kevlar should have much higher tensile strength than fiberglass. I'd like to see an impact test of a panel with fiberglass on the top and Kevlar on the bottom. Then again, I could be wrong.