The issue here, AFAIK, is how long the system waits to send tcp
keepalive probes.  This can be adjusted with the
net.inet.tcp.keepalive sysctl.  You can also adjust the number of
probes sent before the connection is closed via net.inet.tcp.keepcnt.

I also suggest fetching all of the distfiles you need beforehand.
This can be accomplished by running 'make fetch-list | sh' or other
means.


Regards,

--Blair