When and why are conductivities of two neighboring layers too different?

Dear Feflow Masters,
I am trying to simulate a water-table of an area with two layer model, where the lower layer represents tertiary clays (kf 10-7 m/s) and the upper one gravel/sand (kf about 10-3 m/s). The model is set as steady, with unconfined aquifer and unconstrained head. The border conditions (Hydraulic Head and Fluid-Transfer) are defined above the clay (in upper layer).  The lower layer is wavy with relatively steep slopes and I am not able to reach stable simulation. Is this caused by dry cells? I tried to reduce residual water depth to avoid errors in flow directions, as well as the Inflow on the top layer, without any progress. Likewise the vertical refinement of mesh does not lead to any improvement.
Thank you very much for any tips.
PS: for Kf 10-3 m/s for upper and 10-6 m/s for lower layer works the model fine. But the lower layer has to be less conductive.

Try setting slice 1 to free and movable and make slice 2 fixed to prevent complete drying of layer 1. Does this help?

You could also try a pseudo steady state model (steady boundary conditions but allow feflow to resolve equations at a time step of its choosing)...