12. Check whether the sampler behaved
Shows how to tell whether posterior computation can be trusted. You read trace plots, effective sample size, divergences, R-hat, energy diagnostics, and warning messages instead of treating software output as magic.