I also noticed this before.  There is a section in the code where Vortex flow documents something like " I don't think this is the proper way to determine the time step"  I'm not sure if they figured it out later or what, but the note is still there. In general I've gotten better results with a subdivided/odered mesh where the sections look like cubes. seems to solve faster for longer overall simulated time.

I think I put something in the idiots guide about this, but if not, You can get on to the GMSH email exploder and ask how to make an ordered mesh. Also there are some samples on the source folders that are done like that.