Jason Thorpe <thorpej@nas.nasa.gov> writes:

> I must have introduced this bug when I committed an soqinsque() and
> soqremque() optimization from Matt Thomas.  What surprises me is that
> I ran with these changes for months with no problems before committing
> them.

Unless your machine accepts lots of connections you would never see
this.  Also, it appears to happen more with large listen queues...
I have mine set to 128, for instance...

--Michael