Convergence for Initial Interpenetration?

Hi gurus,

How can I improve convergence for initial interpenetration?

My model of a piston seal solves without Large Displacements; but with that enabled (required for hyperelastic material, and I think for correct behaviour) it gives the following failure:

*** A fatal error has occurred. ***
The nonlinear iteration did not converge for the
time value: 0.00000e+00

 o The model may have a buckling mode with a load factor
 smaller than this value. You may determine the smallest
 (linear) buckling load factor by running a buckling
 analysis.

 o The deformations may be large enough that you must
 further refine the mesh. To locate potential problem
 areas, it may help to run an analysis with reduced
 loads and examine the deformed shape of the model.

 o The model may have a static instability.

 o It is possible that the load is too large for the model's 
 material or the model is insufficiently constrained. Check 
 the constraints and material properties, or reduce the maximum load. 

This is for load factor 0 so I can't reduce the load any further!

I've tried reducing the element size, with no improvement.  Any other suggestions?