Pie plots depicting the fraction of CPU h of each of the major calibration and imaging steps for 1.2″-imaging (left panel) and 0.3″-imaging (right panel). The 1.2″ processing was done for one observation with the facet-mode from WSClean in about 7000 CPU h, and the 0.3″-imaging was done for one observation using the predict-subtract method for one observation in about 139 000 CPU h. These numbers scale roughly linearly with the number of observations. The ‘Subtract DDF FoV’ includes the subtraction of sources outside the 2.5° × 2.5° field of view from the last paragraph of Sec. 3.1. The ‘DI VLBI calibration’ includes the DI calibration from Sect. 3.2. The ‘DD VLBI calibration’ includes the DD calibration selection and self-calibration from Sec. 3.3. To highlight the computational costs for imaging compared to the other data reduction steps combined, we indicate in the figure ‘Imaging’ and ‘Other’ in the pie plots.