Notes. All runs assumed a Salpeter IMF with M_top ≤ 150 M_⊙ (see label) and 10% of stars rotating fast enough for chemically homogeneous evolution. The total size of the region formed is 3⋅10⁵ M_⊙ in all cases. For reference, the observed values are Q${}_{\mathrm{He}\mathrm{II}}^{\mathrm{obs}}$ = 1.33⋅10⁵⁰ photons s⁻¹ and L${}_{1550}^{\mathrm{obs}}$ = 2.2–5.5⋅10³⁷ erg s⁻¹. Our results from most runs are able to account for these, regardless of the various assumptions about the star-formation history and thus, the stellar content. The fiducial case mimics the scenario in Paper I (details are given in Sect. 2.3), while the others relax the SFR, lower the M_top, and lengthen the age of the population, meaning later phases of lower-mass models are captured. This is why the source of the stellar C IV emission gets less and less extreme: after the third case, no very massive star (VMS) is needed at all. As for the observed hardness, i.e. the relative intensity of I(4686)/I(H[[INLINE286]]) = A⋅Q(He II)/Q(H I), it should be between 0.02−0.04 observationally, which renders our third population most consistent with measurements (including the C IV line luminosity, which becomes lower with smaller M_top) and thus most favoured, as explained in Sect. 3.2 (also see Fig. 7).