Do you know what versions of software are on the unit? We have seen communications problems with LVRT < 7.0 where the node will not resume communications promptly following a network outage.
Are you able to ping the unit during the time that it is unresponsive? Does the controller's web page come up ok? Can you ftp to the unit? The different services rely on different portions of the TCP stack. In our case ping would always work, FTP and HTTP wouldn't work and MAX would sometimes work. We found that if we waited long enough ( minutes to 2-3 days) the controllers would eventually resume proper communication on their own. A reboot always restored comm immediately.
Moving to LVRT 7.0 made a significant improvement.
We built a test rig with a PC, 3 ethernet hubs and the cFP. The middle hub was power cycled 20S on, 10S off to simulate network problems. With this setup our node communication (VI server) would hang or the node would reboot within 5 to 10 minutes with LVRT 6.1.3, ~ 24 hours or more with 7.0. We have heard that there were significant changes to the LVRT 8.x code and it should be better but we can't go higher than 7.0 right now.
Current versions: LVRT 7.0.0 FP Control Environ 4.1.7 FP VI Mgr 2.0.1 NI-IrDA RT 1.0.0 NI-Serial 2.5.3 NI-Visa 3.0.1 NI-Watchdog 2.1.5
If network issues are causing your problem, you might be able to induce the problem by unplugging and plugging the network cable repeatedly. That would at least give a direction for further investigation.
Another thought would be to set the cFP ethernet settings to fixed values rather than the default auto-negotiate, if it does seem to be network related.