In this work, we explore the behavior of our highly active cathode catalyst NC_Ar + NH3 (labeled NC here) before and after fluorination by F2 at room temperature, in order to deepen our understanding of the activity and stability of this highly active catalyst for oxygen electroreduction in H2/air and H2/O2 PEM fuel cells. We discovered that all Fe-based catalytic sites were poisoned by the reaction with F2, even after 2 min of fluorination, and the H2/O2 fuel cell polarization curves of F2-poisoned NC were the same as the polarization curve of MOF_CNx_Ar + NH3, which is a catalyst devoid of Fe, but having the same CNx active sites as those contributing to the NC activity. The F2-poisoned NC catalysts may be partially (∼70%) reactivated by a heat-treatment at 900 °C in Ar, a temperature at which all Fe–Fm bonds of the catalytic sites and all the C–Fp bonds of the NC carbon support are broken. Based on the deconvolution of the XPS N1s spectra recorded for the pristine and fluorinated NC catalysts, we estimated an average turn-over-frequency (ToF) for the FeN4 catalytic sites in H2/O2 fuel cells. ToF of NC ∼ 0.177 ± 0.020 electrons site−1 s−1 at 0.9 V, which is about five times smaller than the ToF for Pt/C at 0.9 V. The stability of pristine and fluorinated NC catalysts has been studied at 0.6 V in H2/air and H2/O2 fuel cells. The instability behavior of fluorinated NC catalysts was found to be similar to that of MOF_CNx_Ar + NH3, but different from that of NC. It is shown that when the catalyst has active FeN4 sites, it is the model proposed by INRS and using the superposition of two exponential decays that better fits the experimental decay, while the Los Alamos autocatalytic model is preferred when there are no Fe-based active sites in the catalyst or if the FeN4 sites are poisoned (e.g., by fluorination as in this work).