Published arguments, which assign an important role to atomic metastability in the production of ‘narrow’ Zeeman component radiation from the boundary region of fusion plasmas, are examined critically in relation to l-redistribution by proton and electron collisions, and mixing of unperturbed atomic states by the ion microfield and microfield gradient. It is concluded that these important processes indeed severely constrain the contribution from ‘metastable’ states to the generation of the hydrogen Balmer spectra, for electron concentrations above 1012 cm−3, as pointed out before by the present author (Hey et al 1999 J. Phys. B: At. Mol. Opt. Phys. 32 3555). The analysis of collision-induced l-redistribution represents an extension of that used previously (Hey et al 1996 Contrib. Plasma Phys. 36 583), applicable up to higher electron densities. For comparison purposes, we also consider the question of metastability of ionized helium in a low-temperature plasma, and that of some common hydrogenic impurities (C5+ and Ne9+) in a hydrogen (deuterium) fusion plasma. While for low nuclear charge Z the metastability of 2s1/2 levels is quenched by the plasma environment, it is much reduced in high-Z ions owing to the rapid increase with Z of the two-photon electric dipole (2E1) and magnetic dipole (M1) spontaneous transition rates to the ground state, whereas the role of the plasma in these cases is less important. The main new principle elaborated in this work is the sensitivity of atomic line strengths, and hence collision strengths, to perturbation by the plasma environment for transitions between fine-structure sublevels of the same principal quantum number. As the plasma microfield strength grows, ‘allowed’ transitions diminish in strength, while ‘forbidden’ transitions grow. However, owing to violation of the parity selection rule, there is an overall loss of collision strength available to transitions, resulting from the appearance of significant ‘self-strength’ contributions, in accord with the sum rules for the line strengths, which remain valid over the range of fields considered. Thus, the relative effectiveness per perturber of both electron and ion collisions, for inducing population transfer between fine-structure sublevels, diminishes as the sublevels evolve from a fine-structure dominated to a Stark-effect-dominated regime. In the concluding discussion, we mention that this finding may have a bearing on discrepancies claimed between Stark broadening theory developed by Griem (1967 Astrophys. J. 148 547) and by Watson (2006 J. Phys.B: At. Mol. Opt. Phys. 39 1889), and the measurements of Bell and co-workers (2000 Publ. Astron. Soc. Pac. 112 1236; 2011 Astrophys. Space Sci. 335 451) for high-n radio recombination lines from galactic H II regions. In the absence of detailed modelling to test this suggestion, however, it would be premature to attempt to draw any firm conclusions along these lines.