We reassess the viability of a cosmological model including a fourth additional sterile neutrino species that self-interacts through a new pseudoscalar degree of freedom. To that end, we perform a series of extensive analyses fitting various combinations of cosmic microwave background (CMB) data from Planck, the Atacama Cosmology Telescope (ACT) and the South Pole Telescope (SPT), both alone and in combination with Baryon Acoustic Oscillation (BAO) and Supernova Ia (SnIa) observations. We show that the scenario under study, although capable to resolve the Hubble tension without worsening the so-called S 8 tension about the growth of structures, is severely constrained by high-multipole polarization data from both Planck and SPT. Intriguingly, when trading Planck TE-EE data for those from ACT, we find a 3σ preference for a non-zero sterile neutrino mass, m s = 3.6 +1.1 −0.6 eV (68% C.L.), compatible with the range suggested by longstanding short-baseline (SBL) anomalies in neutrino oscillation experiments. The preference is mostly driven by ACT favouring a higher value for the primordial spectral index n s with respect to Planck and SPT. The pseudoscalar model appears indeed to provide a better fit to ACT data with respect to ΛCDM (∆χ 2 −5). In particular, we show that the mild tension between Planck and ACT is due to the different pattern in the TE and EE power spectra on multipoles between 350 1000. We also check the impact of marginalizing over the gravitational lensing information in Planck data, showing that the model does not solve the CMB lensing anomaly. Future work including higher precision data from current and upcoming CMB ground-based experiments will be crucial to test these results. Contents 1 Introduction 1 2 The pseudoscalar sterile neutrino self-interaction model 4 3 Methods and data 4 4 Results 5 4.1 Planck + A TTTEEE L + A φφ L 6 4.2 Planck TT + SPT 8 4.3 Planck TT + ACT 9 4.4 Best-fit cosmology and its impact on the CMB 10 5 Planck vs ACT: a deeper look 13 6 Comparison with SBL neutrino oscillation experiments 16 7 Conclusions 18 A Best-fit parameter values and χ 2 per experiment 20 B Additional triangle plots 22 C Impact on the linear matter power spectrum 22 Pseudoscalar model Parameter ACT +BAO+SnIa Planck +ACT +BAO+SnIa 100 ω b 2.185 +0.