Context. Recent James Webb Space Telescope (JWST) spectroscopic observations of the z = 10.6 galaxy GN-z11 have revealed a very peculiar UV spectrum exhibiting intense emission lines of nitrogen, which are not typically detected in galaxy spectra. This observation indicates a super-solar N/O abundance ratio at low metallicity, which only resembles the abundances seen in globular cluster (GC) stars. This discovery suggests that we might be seeing proto-GCs in formation or possibly even signatures of supermassive stars. Aims. To examine whether other objects with strong N IV and/or N III emission lines (N-emitters, hereafter) exist and to better understand their origin and nature, we have examined the available JWST spectra and data from the literature. Methods. Using the NIRSpec/JWST observations from CEERS, we found an extreme N-emitter, CEERS-1019 at z = 8.6782, showing intense N IV ] λ 1486 and N III ] λ 1750 emission. From the observed rest-UV and optical lines, we conclude that it is compatible with photoionization from stars and we have determined accurate abundances for C, N, O, and Ne, relative to H. We also (re-)analyzed other N-emitters from the literature, including three lensed objects at z = 2.3 − 3.5 (Sunburst cluster, SMACS2031, and Lynx arc) and a low-redshift compact galaxy, Mrk 996. We carried out a comparison among the observed abundance ratios to observations from normal star-forming galaxies, predicted wind yields from massive stars, and predictions from supermassive stars (SMS with ∼10 4 − 10 5 M ⊙ ). Results. For CEERS-1019, we find a highly supersolar ratio log(N/O)= − 0.18 ± 0.11, and abundances of log(C/O)= − 0.75 ± 0.11 and log(Ne/O)= − 0.63 ± 0.07, which are normal compared to other galaxies at the low metallicity (12 + log(O/H) = 7.70 ± 0.18) of this galaxy. The three lensed N-emitters also show strongly enhanced N/O ratios and two of them normal C/O. The high N/O abundances can be reproduced by massive star winds assuming a special timing and essentially no dilution with the ambient interstellar medium (ISM). Alternatively, these N/O ratios can be explained by mixing the ejecta of SMS with comparable amounts of unenriched ISM. Massive star ejecta (from WR stars) are needed to explain the galaxies with enhanced C/O (Lynx arc, Mrk 996). On the other hand, a SMS in the “conveyer-belt model” (put forward to explain globular clusters) would predict a high N/O and small changes in C/O, compatible with CEERS-1019, the Sunburst cluster, SMACS2031, and GN-z11. Based on the chemical abundances, possible enrichment scenarios, and other properties (e.g., their compactness and high ISM density), we discuss which objects could contain proto-GCs. We suggest that this is the case for CEERS-1019, SMACS2031, and the Sunburst cluster. Enrichment in the Lynx arc and Mrk 996 is likely due to normal massive stars (WR), which implies that the star-forming regions in these objects cannot become GCs. Finally, we propose that some N-emitters enriched by SMS could also have formed intermediate mass black holes and we suggest that this might be the case for GN-z11. Conclusions. Our observations and analysis reinforce the suggested link between some N-emitters and proto-GC formation, which is supported both by empirical evidence and quantitative models. Furthermore, the observations provide possible evidence for the presence of supermassive stars in the early Universe ( z > 8) and at z ∼ 2 − 3. Our analysis also suggests that the origin and nature of the N-emitters is diverse, including objects such as GN-z11, which may possibly host an active galactic nucleus (AGN).