minicom works conserver randomly doesn't

We have a linux cluster running conserver on 8 management nodes, each 
managing 32 consoles with rocketport cards.  This has been working for 
quite a while.  However, just recently, 3 of these nodes have blocks of 
consoles that are marked as being down with a console -u.  The consoles 
are accessible and fully working through minicom, and the rocketport 
boxes have completely passed diagnostics.  Each line of the conserver.cf 
looks something like this:

node1:/dev/ttyR4:9600p:/disks/space0/logs/node1.log:

On one machine all ports are marked as down, on another it's 16 of 32, 
and on another it's 19/32, but each machine consistently has the same 
ones marked as down between conserver restarts/reboots/etc.

If i do "console nodename" I get:
console: nodename is down

^eco says:
[line to host is down]

I'm out of ideas.  Has anyone seen this before?

Joe