Histograms showing the number of likelihood evaluations per second (top), the final accepted fraction of evaluations (middle), and the total forecast run-time in hours (bottom) of our stage III forecasts, coloured according to the probe configurations (as in previous figures, and also given in the legend; see Table 2). We note that these are indicative estimates for runtimes, excluding some forecasts that were rerun at later stages, and including some variability stemming from the number of parallel threads utilized for each specific forecast. As a result, some forecast timings were unrepresentative (< 10 evaluations s⁻¹), and these are not shown.