This paper reports results of numerical simulations of stratified tidal flow over the Knight Inlet sill. A non–hydrostatic, two–dimensional model is used, which incorporates a no–slip bottom boundary condition through the use of a vertical eddy viscosity/diffusivity parametrization that is non–zero only near the bottom. In inviscid model runs, a large lee wave is rapidly formed, which quickly breaks, leading to the formation of a high–drag state and a strong downslope jet in the early stages of the ebb tide. The use of a no–slip bottom boundary condition results in boundary–layer separation from near the top of the sill. This significantly reduces the amplitude of the lee wave during the initial stages of the flow development. For most model runs, a large lee wave is ultimately formed and the separation point moves down the lee of the sill to a position immediately downstream of the lee wave. The transition to this high–drag state is significantly delayed compared with inviscid model runs. Weakened stratification immediately above the sill, inclusion of an eddy viscosity/diffusivity above the bottom and a pool of dense water on the downstream (seaward) side of the sill can all contribute to a delay in the transition to a high–drag state, and can eliminate it entirely. For one model run using a vertical eddy viscosity parametrization above the bottom, a reduction of the vertical diffusivity eliminated the formation of a high–drag state. This suggests that at least in some cases entrainment into the lee wave can cause its growth and result in the formation of a high–drag state.