Title: The progenitor galaxies of stellar haloes as "failed" Milky Ways

Abstract: The stellar halo of the Milky Way records the history of its interactions with dwarf galaxies, whose subsequent destruction results in the formation of an extended stellar component. Recent works have suggested that galaxies with masses comparable to the Large Magellanic Cloud (LMC, $M_\star \sim 10^9\,{\rm M}_\odot$) may be the primary building blocks of the stellar halo of our Galaxy. We use cosmological simulations of the $\Lambda$ Cold Dark Matter model to investigate LMC-mass galaxies at $z=1-2$ using a semi-analytic model of galaxy formation. We find that LMC analogues at $z=2$ evolve until the present day along three distinct pathways: (1) those that are destroyed in Milky Way-mass hosts; (2) those that are themselves the main progenitors of Milky Way-mass galaxies; and (3) those that survive until $z=0$, with stellar mass $\sim$1.0 dex lower than typical Milky Ways. We predict that the properties of these galaxies at $z=2$ (stellar metallicities, sizes, gas content etc.) are largely indistinguishable, irrespective of which of these pathways is eventually taken; a survey targeting such galaxies in this redshift range would struggle to tell apart a 'destroyed' stellar halo progenitor from a 'surviving' LMC analogue. The only factor that determines the eventual fate of these galaxies is their proximity to a neighbouring Milky Way main progenitor at $z=2$: while the mean separation to a 'surviving' galaxy is around 7 Mpc, it is only 670 kpc to a 'destroyed' galaxy. This suggests that old stellar populations in the Milky Way may share intrinsic (i.e. non-dynamical) properties that are essentially indistinguishable from progenitors of its stellar halo.