Convergence issues with Shear thinning fluids
 

I am trying to model solid suspension in non-newtonian systems in stirred tanks. I am using steady state simulations. Impeller motion is handled by MRF. I have a granular eulerian phase for the solid.

My issue is with the herschel-bulkley viscosity. When I added significant shear thinning properties, I get significant convergence issues. With a power law constant of .99, my residuals look fine, but when I lower it to .9, my residuals will spike up for a couple hundred iterations, then start to descend, then spike back up again. The phase and velocity distributions are terrible too.

Any Ideas what might cause this?