We performed a series of numerical experiments to quantify the sensitivity of the predictions for weak lensing statistics obtained in raytracing DM-only simulations, to two hyper-parameters that influence the accuracy as well as the computational cost of the predictions: the thickness of the lens planes used to build past light-cones and the mass resolution of the underlying DM simulation. The statistics considered are the power spectrum and a series of non-Gaussian observables, including the one-point probability density function, lensing peaks, and Minkowski functionals. Counter-intuitively, we find that using thin lens planes ($< 60~h^{-1}$Mpc on a $240~h^{-1}$Mpc simulation box) suppresses the power spectrum over a broad range of scales beyond what would be acceptable for an LSST-type survey. A mass resolution of $7.2\times 10^{11}~h^{-1}\,M_{\odot}$ per DM particle (or 256$^3$ particles in a ($240~h^{-1}$Mpc)$^3$ box) is sufficient to extract information using the power spectrum and non-Gaussian statistics from weak lensing data at angular scales down to 1 arcmin with LSST-like levels of shape noise.