Hi Matt, I've been thinking about the need for us to create some special verification datasets where we know what the right answer is, so we can test the various routines, like the streamline plot, to see if we get the expected plot. These should be conceptually simple datasets (like a left-handed vortex of vectors) where we know what the result should look like.
I had tried something earlier with a volume render dataset when I was playing with left/right handedness. In that case, I had simply made my dataset in NumPy as a cubic array and then fed it into yt -- that seems to be an easy route to testing these things.
Mike