Messed-up positioning on metaball parented to a mesh with vertex duplication (2.71)


First-time poster and recent Blender convert.

I’ve been intrigued by the possibility of parenting a metaball to meshes and then activating vertex duplication on the latter.

Unfortunately, in practice there appears to be a significant bug in the 3D positioning of the dupli metaballs when this is done.

If any metaball is parented to any mesh (e.g. torus, icosphere), and vertex duplication is enabled on the mesh, even with both objects at (0,0,0) and no rotation, the metaball duplis are offset from the origin. They ARE in the correct configuration, but are centred at some arbitrary (x,y,z). Location as reported in the object data of both objects is still 0,0,0.

To make matters more frustrating, any interaction with the viewport (e.g. zoom, pan) superficially corrects the issue, but the moment I switch to rendered view, or indeed make a real render, the problem is back again. In fact, various interactions appear to aggravate or temporarily solve the problem at random!

It’s possible of course to make the duplis real and proceed from there, but that precludes the very interesting possibilities of adding dynamic displacements/warps/waves etc. to the dupli mesh (when I try this with the offset problem present, the centre of the offset duplis wobbles around as the displacement/wave/warp evolves).

Has anyone seen this/can anyone reproduce it? (Using 2.71 here) Any ideas for a fix or a workaround?

Many thanks.