4 worker_connections is not enough while accepting new connection

[Roxis]

Здравствуйте уважаемые,

С worker_connections 4 работают параллельно только 2 коннекта.
Хотелось бы узнать, куда деваются первых 2 коннекта и нет никакой очереди.

# nginx.conf
user www;
worker_processes 1;

events {
    worker_connections 4;
}

http {
error_log   logs/error.log  debug;

server {
    listen  127.0.0.1:80;
    root    /www;
}

}

# uname -srm
Linux 2.6.21.3 i686

# error.log
2007/06/02 18:03:39 [debug] 10187#0: accept on 127.0.0.1:80, ready: 0
2007/06/02 18:03:39 [debug] 10187#0: malloc: 080B1F78:256
2007/06/02 18:03:39 [debug] 10187#0: *3 accept: 127.0.0.1 fd:5
2007/06/02 18:03:39 [debug] 10187#0: *3 event timer add: 5: 60000:3979240381
2007/06/02 18:03:39 [debug] 10187#0: *3 epoll add event: fd:5 op:1 ev:80000001
2007/06/02 18:03:43 [debug] 10187#0: accept on 127.0.0.1:80, ready: 0
2007/06/02 18:03:43 [debug] 10187#0: malloc: 080B2080:256
2007/06/02 18:03:43 [debug] 10187#0: *4 accept: 127.0.0.1 fd:8
2007/06/02 18:03:43 [debug] 10187#0: *4 event timer add: 8: 60000:3979244103
2007/06/02 18:03:43 [debug] 10187#0: *4 epoll add event: fd:8 op:1 ev:80000001
2007/06/02 18:03:47 [debug] 10187#0: accept on 127.0.0.1:80, ready: 0
2007/06/02 18:03:47 [alert] 10187#0: 4 worker_connections is not enough while 
accepting new connection on 127.0.0.1:80