Diagnostic plots of the Hamiltonian MCMC sampler with KiDS-1000 data and our final analysis set-up for fixed projection parameters as in Table 1. Top panel: Example of f_δ, mn variations along one random chain at the beginning of the burn-in, starting from the random initial point i = 0 (blue line). The orange line depicts the (negative log-) posterior probability −2lnP(π_i|d). Bottom panel: Decrease in the correlation of f_δ, mn and f_δ, kl with δi iterations lag as estimated in ten randomly chosen, independent chains. The orange lines follows, one line per chain, the maximum absolute-value Pearson correlation coefficient |R_mn, kl|, while the blue lines follows the average |R_mn, kl| over all variable combinations (mn, kl). The sampling points quickly decorrelate after a few hundred iterations; the flattening after lag δi ∼ 100 is due to the statistical error in the estimated |R| for chains of finite length n_mcmc = 5 × 10⁴.