Goodness of fit sampling distributions. Shown are histograms of minimum χ²-values obtained from 1000 mock realisations of data vectors analysed with our cosmic shear and GGL analysis pipeline. Blue points correspond to our baseline setup; red points to the result using a sabotaged data vector computed with the S3 and S4 redshift distributions systematically shifted upwards by 5σ of their Gaussian prior width. The curves are best fits of a χ²-distribution with its mean as free parameter. The solid blue curve is a good fit to the baseline results (p-value of 0.16), while the sabotaged distribution is still well fitted by a χ²-distribution (red dashed curve; p = 0.53) which however is incompatible with the baseline (p < 10⁻⁵).