Terminal velocity v_∞ (top) and the ratio v_∞/v_esc,eff (bottom) against log L/L_⊙ (left) and T_eff (right) for the R136 sample (solid dark blue diamonds) compared to predicted values (yellow squares, red stars, and green pentagons Björklund et al. 2021; Krtička & Kubát 2018 and Vink & Sander 2021, respectively). Light purple triangles denote R136 sources for which we could not derive v_∞ values (see Sect. 4.3), and in the bottom plot also includes the WNh stars, where we cannot be too confident about v_esc,eff. Grey circles around the points of the WNh stars allows to distinguish them from the O-type stars. Dark and light coloured error bars denote 1 and 2σ uncertainties, respectively. The lines, plotted in different styles (see legend) show linear fits to both the observed (dark blue) and predicted values (red, orange and green). In the bottom panel, the black dashed line shows the empirically derived v_∞/v_esc,eff = 2.6 (Lamers et al. 1995). The grey shaded regions in the luminosity plots correspond to those in Fig. 12. We note that we do not show any points of models with T_eff = 35 000 K for the Vink & Sander (2021) predictions; in this regime the predicted terminal velocities largely exceed the velocity scale of this plot. The linear fit through the Vink & Sander (2021) predictions also excludes these points.