SN 2019qem: decades of extreme mass loss before the explosion of a luminous, long-lived Type IIn supernova

We present the luminous, long-lived Type IIn SN 2019qem and use it to reconstruct how much mass its progenitor lost and when. We combine host-subtracted optical and mid-infrared photometry and 26 optical spectra from -22 to +699 d. We model the bolometric light curve with thin-shell CSM-interaction models in REDBACK, with the shock velocity constrained by the H$α$ line widths, and compare them with an analytic interaction model. SN 2019qem rose in $\approx30$d to $M_I = -20.5$mag and a bolometric peak of $7.0\times10^{43}$ erg s$^{-1}$, and radiated $6.9\times10^{50}$ erg in the optical within 575 d, closely resembling luminous, strongly interacting SNe IIn such as SN 2010jl and SN 2015da. A Heii/Niii flash feature at -22 d and unresolved narrow Balmer lines reveal dense, slow CSM. From +78 to +445 d the H$α$ wings are shaped by electron scattering, and the line stays blueshifted by 400-700 km s$^{-1}$, which points to an optically thick, asymmetric interaction region rather than to newly formed dust. In the preferred model, $\approx7 M_\odot$ of ejecta run into $\approx9 M_\odot$ of gas ejected $\approx2$ yr before explosion and then into a wind of $\approx0.6 M_\odot$ yr$^{-1}$ (for $v_w = 100$ km s$^{-1}$), sweeping up $\approx32 M_\odot$ by +544 d. The mid-infrared excess brightens for two years while the optical fades, tracing dust heated by the ongoing interaction. The host is a low-mass, star-forming galaxy, $\log(M_*/M_\odot) = 9.19$, with a sub-solar metallicity at the explosion site. In its final decades the progenitor of SN 2019qem lost mass far faster than any steady stellar wind, ending with an eruption of $\approx9 M_\odot$ two years before core collapse, consistent with luminous-blue-variable-like eruptive mass loss. Taken at face value, the mass budget implies an initial progenitor mass of $\gtrsim40 M_\odot$.