Kind of OT, but losing data in these cases is not even what I'm most concerned about with ES: I had to recreate the ES cluster way too many times now that I'm really glad that PostgreSQL keeps on running, or just restarts without bricking the data files, even after OOM or out of disk errors.