Complex Initrd Issue Probably Involving A Broken Sh

We had someone take a look at our server for security issues, and they believed that one way of keeping "hackers at bay" would be to write an sh wrapper that only allowed root access. Well, this immediately broke the system.

I have spent about six hours now attempting to fix this. We're using Red Hat Enterprise 4, and the recovery disc doesn't have the SCSI drivers we need to even mount the hard drive, so that's out. I can't skip initrd because it apparently contains the drivers for the hard drive, but it's also not allowing the system to boot up. I have tried setting kernel emergency in grub, but it still loads initrd and breaks. Without initrd, I get a kernel panic while mounting the hard drive. Annoyingly enough, doing cat initrd-version.img at the grub prompt shows the file just fine, so why is initrd required for mounting the drive if it's already accessible? Also, it doesn't look like there's any way to edit initrd from grub, which is a real pain in the head because I could always comment out the offending lines. And of course, even though I can see the initrd file, I can't do anything else with the file system. I have requested of the data center to provide a boot disc that can mount the hard drive, and if that can be done, then I can go in and make the changes immediately and hope that it works, but without that, it seems I'm stuck. Any suggestions?

When I say "broken", what happens is I get "Out of memory: Killed process x (procname)" which then repeats itself infinitely, killing any process which would load.

The data center says that they HAVE no boot discs capable of mounting the hard drive. This is utterly ridiculous, because the only option from here is to order the expensive OS reload. I'll never use Red Hat again.