Some updates in the software image require recreating the initrd. Bright Cluster Manager is not aware of such updates. As result, when regular nodes are rebooted, the old initrd in the software image may conflict with the kernel/module updates and cause a hang at the time of loading or maybe later.
The fix in that case is to regenerate an initrd using cmsh's createramdisk command (from the softwareimage mode), or cmgui's "Recreate Initrd" command (from the "Software Images" resource, in the "Kernel Config" tab). Then reboot the node.
This case and other possible node boot problems are discussed in the section of the Admin manual on "Troubleshooting The Node Boot Process"