Okay, so I thought that perhaps the size of the map influenced whether this happened or not, and it does seem to do that, though not due to the map size per se.
I've been watching the 2 threads for I/O and the send is fairly steady at 1.5 kBps. Looks good.
The receive thread looks like a sine wave though. At the beginning and end of a round, it starts around 2 kBps, give or take 500 Bps. It will then begin to climb to anywhere from 8-10 kBPS until half of the players are dead, then decline again. How quickly it climbs and how high seems to depend upon engagement. The more people on the server, and the faster they engage, the quicker and higher this figure climbs.
Now I'm guessing that this makes sense, but I can reliably tell you when half the population of the server is dead; and when the round ends, I can tell you approximately how many of the winning team survived by the peak and valley of the traffic.
So how this plays out in game:
A large map is actually better for this issue. The reason being is that you have such varied speed between cavalry and athletics and armor, that the engagement begins as a trickle, peaks when about half the players die, and trickles off until round end as there are fewer engagements. With this trickle effect, you only build up to around 10-11 kBps, even with 80+ on the server.
A smaller map it plays out differently. Rather than the engagements being spread out, everyone rapidly is in combat. This means that instead of a trickle of information, suddenly there is a flood as everyone is engaged at once. That is when I noticed this issue the most. The highest peak I saw was over 16 kBps, and I suspect that somewhere between 11 and 16 kBps is where things suddenly begin to choke and we experience this issue.
I supposed I could load up The Arena as that is the smallest map that I know of, and should be more pronounced than any other.
Let me know what you think of the theory and if there is anything that I can look at for you.
