Thanks guys! It was very helpful to see those other replays.
The disconnection was definitely not recorded in the host's (avirex's) replay, but was recorded fine in the clients' replays. I thought I had fixed this bug in v3.6.30.0, but apparently it can still happen under certain circumstances. I think I have an idea of what might be causing it.
Did it take about 150 seconds for Anubis to time out? (This question is for everyone except Anubis — to Anubis: since the overhaul to the disconnection code in v3.6.30.0, timeout of a host from a client's perspective is not yet implemented... it just keeps waiting forever, so you have to quit manually... which I can see you did from your replay.)