Turbulent motions regulate vertical transport of momentum, heat, moisture and pollutants in the atmospheric boundary layer. From a numerical perspective, modeling such motions becomes challenging at kilometer and sub-kilometer resolutions, as the horizontal grid spacing of the model approaches the size of the most energetic convective eddies in the boundary layer. In this range of resolutions, typically termed ‘terra incognita’ or ‘gray zone’, partially resolved convective structures are grid-dependent and neither traditional 1D mesoscale parametrizations nor 3D closures typical of Large Eddy Simulations are theoretically appropriate. However, accurate numerical modeling at gray zone resolutions is a key aspect in several practical applications, such as proper coupling of mesoscale and microscale simulations. While some progress has been achieved in recent years through idealized simulations and theoretical considerations, the evaluation of different approaches in real convective boundary layers (CBL) is still very limited. Leveraging on a new set of one-way nested, full-physics multiscale numerical experiments, we quantify the magnitude of the errors introduced at gray zone resolutions and provide new perspectives on recently proposed modeling approaches. The new set of experiments is forced by real time-varying boundary conditions, spans a wide range of scales and includes traditional 1D schemes, 3D closures, scale-aware parametrizations and strategies to suppress resolved convection at gray zone resolutions. The study area is Riyadh (Saudi Arabia), where deep CBLs develop owing to strong convective conditions. Detailed analyses of our experiments, including validation with radiosonde data, calculations of spectral characteristics and partitioning of turbulent fluxes between resolved and subgrid scales, show that (i) grid-dependent convective structures entail minor impacts on first order statistics of the flow due to some degree of ‘implicit scale-awareness’ of 1D parametrizations and (ii) 3D closures outperform traditional and scale-aware 1D schemes especially in the surface layer, among other findings. The new simulation suite provides a benchmark of real simulations that can be extended to assess how new techniques for simulations at gray zone resolutions perform.