Ok, I know what is going on here. Basically one of the AStarPathfinderParams members is called 'uMaxNoOfNodes' and this basically controls the maximum number of cells that the algorithm will explore (I think this was to stop it getting stuck in a never ending loop). It defaults to 10000, and changing it to 300000 does seem to let the algorithm find the path between the two corners of your cube.
Code:
AStarPathfinderParams<SimpleVolume, MaterialDensityPair44> params(&volData, Vector3DInt32(0,0,0), Vector3DInt32(63,63,63), &listResult, 1.0f, 300000);
However, it is
extremely slow. I took somewhere between 5 minutes and 1 hour to process on my laptop (not sure exactly as I went off for dinner). In other words you can't really use it in it's current state.
I'm sure there are a lot of optimization opportunities but this is something I'll be looking into myself as I'm not using pathfinding at the moment. Alternatively you can try running the pathfinding on a smaller version of you volume.
I'm not actually sure what the expected performance of A* is. In terms of the number of elements a 64x64x64 volume is the same as a 512x512 image, but I don't know whether A* would normally work on an image that size as I'm not an AI guy. But even with the same number of nodes, the volume case is harder because it needs to explore in three dimensions instead of two.
So there's probably some slow code in there, but it probably a tough problem as well.