manager machine uptime hiccups?

after upgrading from backburner 2.3 to 3.0.2, our once rock-solid manager 
application has been displaying stability issues.

after a few days of uptime, the manager application seems to "hiccup"
and 
shutdown momentarily--render nodes display connection errors and report that 
they are "assuming the manager is down".  "could not find a
manager" 
warnings appear, and soon after "Error recieving information: an existing 
connection was forcibly closed by the remote host. (0x2746)." appears and
repeats 
many times over. 

eventually the manager application comes back online, and render nodes 
register fine.

anyone know what would cause this hiccup?  why 3.0.2 would demonstrate these 
symptoms but 2.3 didn't?

thanks for any info..

greg