When members go down (which is expected to happen at some point) there are very verbose error messages, see included .txt file for a 3-node replica set, being shut down one node at a time. Maybe no stack trace by default, or after the first error?