Well, so last night, after getting back from a birthday party, I thought to myself; does the marching cube algorithm produce manifold meshes? And the answer is “NO”!
The original algorithm can very easily produce holes, an improvement of that algorithm called the M33 followed a few years after and it added 33 additional cases to cover the ambiguities in the original one. Then around 2002, another improvement was published that had one known problem, and now in 2013, a paper and contributions to the open source implementation claims that they have it all covered to generate a manifold mesh.
So, to just throw fixed points in a grid at the marching cubes algorithms should produce good meshes, if and only if you use the C-M33 version of it, which is the correction of the correction that is built on the extended version of the original one. :p
Otherwise, I worked mostly through integrating my native library into my application over the weekend as well. Have not cleared up all the errors yet so I can test it.
V 