Multi-messenger observations will eventually solve this, e.g. a correlation with LIGO, if the sensitivity is good enough and they're 'low mass' collisions. The

Multi-messenger observations will eventually solve this, e.g. a correlation with LIGO, if the sensitivity is good enough and they're 'low mass' collisions. The energy needed for these signals already narrows down the category of objects it could be. As the article says, neutron star collisions and mergers, magnetars, and of course black holes.