Solution diverges when mesh is refined
 

I am trying to do a saturated pool boiling simulation on a finned surface using ANSYF Fluent VOF Lee evaporation-condensation model.

I got some initial results without problem with coarse mesh. But, when I refined the mesh to 2 times, the solution diverges at initial time steps itself.

I tried reducing the time step (up to 10^-9) and tried first order as well as second order schemes. Also tried switching off turbulence model.

I can't understand, how its perfectly working for coarse mesh and blows up for finer mesh with same setup?

The pictures for coarse and fine mesh are attached. The snapshot of how resudual are shooting up is attached.

10^-9 seems far too small. It often helps to increase the time step size at the beginning of a simulation instead of decreasing it.