Weird And Yet Ironic Rebooting

I have compiled a RT kernel using 2.6.24 and it has been working great. All of our machines are runnin Centos 5. No issues like the previous configurations on Dual Quad Xeons 5335's and 8GB of RAM. As well as one box with Dual Core Xeon 5140's with 4GB of RAM. They have identical motherboards. My issue is that when you do a soft reboot, it doesn't "technically" reboot. It does, but it doesn't. Let me explain. It kills my ssh connection and you can't ping it for maybe say a minute or 2. When it comes back up, you get a new SSH key to login. You hit ok, and you try logging in. It doesn't accept any passwords. The server basically becomes inoperable.

The only way to solve this is to do a hard reboot. Phsyically rebooting the server.
What could be causing this? This only started to happened when using the RT kernel. Is there an option that may be conflicting with it, however our 3rd and 4th box, with the same configuration has no issues like this. Soft reboots work great.

Here is the message log.

Feb 19 04:00:02 server530 shutdown[25342]: shutting down for system reboot
Feb 19 04:00:02 server530 init: Switching to runlevel: 6
Feb 19 04:00:03 server530 smartd[3037]: smartd received signal 15: Terminated
Feb 19 04:00:03 server530 smartd[3037]: smartd is exiting (exit status 0)
Feb 19 04:00:09 server530 ntpd[20917]: ntpd exiting on signal 15
Feb 19 04:00:09 server530 rpc.statd[2570]: Caught signal 15, un-registering and exiting.
Feb 19 04:00:09 server530 portmap[25711]: connect from 127.0.0.1 to unset(status): request from unprivileged port
Feb 19 04:00:10 server530 auditd[2469]: The audit daemon is exiting.
Feb 19 04:00:10 server530 kernel: audit(1203415210.158:8204): audit_pid=0 old=2469 by auid=4294967295
Feb 19 04:00:10 server530 pcscd: pcscdaemon.c:529ignal_trap() Preparing for suicide
Feb 19 04:00:10 server530 pcscd: hotplug_libusb.c:361:HPEstablishUSBNotifications() Hotplug stopped
Feb 19 04:00:11 server530 pcscd: readerfactory.c:1350:RFCleanupReaders() entering cleaning function
Feb 19 04:00:11 server530 pcscd: pcscdaemon.c:489:at_exit() cleaning /var/run
Feb 19 04:00:11 server530 kernel: Kernel logging (proc) stopped.
Feb 19 04:00:11 server530 kernel: Kernel log daemon terminating.
Feb 19 04:00:12 server530 exiting on signal 15

Oh and this is to reboot the server nightly at 4am, when this issue occurs.

0 4 * * * /usr/bin/reboot > /dev/null 2>&1