Toward a complete understanding of the properties of GW190521
Interpreting GW190521 remains challenging due to its exceptionally high total mass and correspondingly small number of observable gravitational-wave cycles, which gives rise to degeneracies between different physical interpretations. Motivated by the outstanding questions surrounding the event, we present a comprehensive reanalysis of GW190521 using waveform models drawn from multiple families, including a model that includes both eccentricity and spin-precession. We find evidence that the source was highly eccentric, with an inferred eccentricity of $0.81^{+0.05}_{-0.01}$ at 5 Hz, and show that non-eccentric interpretations arise from failing to probe the high-likelihood region at large eccentricities. We further demonstrate that when eccentricity is included in the analysis, there is no indication of spin precession, suggesting that the apparent precession in quasicircular analyses is at least partially driven by an eccentricity--spin-precession degeneracy. While all waveform models favor a comparable-mass binary, one model admits a secondary mode at more asymmetric masses. We find a robust probability $(\geq97\%)$ that the primary black hole lies within the pair-instability supernova mass gap ($\sim60-130\,M_\odot$), whereas the probability that the secondary lies in the gap depends on the waveform model. Through inference of GW190521-like injections, we show that the waveform systematics observed in the analysis of GW190521 could arise from the response of models to unmodeled eccentricity, rather than from intrinsic differences between waveform models. When analyzing a quasicircular spin-precessing injection similar to GW190521, we find no significant model dependence in the inferred parameters. Finally, because the inferred eccentricity lies beyond the calibration regime of current models, we test its robustness by recovering highly eccentric numerical-relativity waveforms.