Fun story: after the last report which called them out for not talking about write loss by default, MongoDB updated their Jepsen page to say that the analysis didn't observe lost writes. I guess they assumed that people wouldn't... read the abstract? Let alone the report?