Networking :: Debian Lenny 5.0.7 AMD 64 Connection Stops Running?

I have a Dell PowerEdge 2900 on which I just installed from scratch Debian Lenny 5.0.7 AMD 64bit. I am having issues where when I ifup a second nic card the server will maintain connectivity for a period then all connectivity is lost and I have to ifdown all eth interfaces and bring them back up to regain connectivity. Here's how the server is laid out: I have 12gb or ram with 2 Xeon 5450s in the server, as I mentioned it's a Dell Poweredge 2900. It has a build in Perc6i raid controller with 2 drives in Raid 1 for the OS and 4 drives in Raid 5 for storage. It also has a Perc 5e card in it connected to 3x6TB raid arrays. I mention this because they are all sharing the PCI bus if that matters.

The nic cards I started with were the two builtin Broadcoms and lspci reports them both as:

Broadcom Corporation NetXtreme II BCM5708 Gigabit Ethernet (rev 12)

After I started experiencing the problem I added an inexpensive Trendnet based off the Realtek RTL-8169 chip rev(10). Debian sees this card without any need for additional drivers. The Broadcoms required the non-free network drivers provided by and built against the Debian kernel (no custom drivers). One nic card is configured with a public IP address and I use IP tables to reign in the traffic to http and https only. There is also an outbound openvpn tunnel from this machine back to our offices through this interface. The second nic card is hard wired to a mac mini directly and is assigned an 10.x.x.x address. I am able to see the mac mini from it.

The amount of time the network connection stays up doesn't seem to follow any set pattern. I have flood pinged the 10.x.x.x while running apt-get and one time it fails, the next time it doesn't. Sometimes the network connection degrades and I can do an nslookup one second and get nothing and the next I get a response. Finally though it fails and it requires me to ifdown the eth connections and ifup them to get them going again. I don't really see anything in the logs. I've looked in kern.log, messages, syslog and pretty much came up empty, but not 100% certain I'm looking for anything more subtle than obvious errors. The routes all look good too.

Naturally, as Murphy's Law would have it, this is a remote server at a colo so that doesn't make troubleshooting any easier. What I have done to temporarily prevent me from losing complete access to the server is setup a cron script that runs once a minute and pings something. If there is 100% packet loss, it restarts the nics and I'm going again for another arbitrary period of time. Lastly, this has no effect on the OS running. I don't have a hung system when this happens just a trashed network connection.