negative tcp_tw_count and other TIME_WAIT weirdness?

[John Salmon]

I have several fairly busy servers reporting a negative value
for tcp_tw_count.  For example:

bash-2.05a# cat /proc/net/sockstat
sockets: used 121
TCP: inuse 50 orphan 0 tw -65048 alloc 81 mem 26
UDP: inuse 15
RAW: inuse 1
FRAG: inuse 0 memory 0
bash-2.05a# 

When I look at netstat -n, I see many (hundreds) connections 
stuck in TIME_WAIT.  They've been there for at least a few hours,
and probably much longer (days).

Is this expected behavior?  A known bug?

FWIW, I'm using a RedHat kernel, 2.4.18-24.7.xsmp on a 2-processor Athlon
system.   If this looks like a bug I'll try to reproduce it with
an unmodified kernel.

Thanks,
John Salmon