This is a bit like asking for the precise cause of death when someone takes a direct hit from a nuclear weapon.

With no /dev/tty to write on we will never know. The executable will never get far enough to detect a missing shared library since there is no /dev/zero or /dev/full left to permit the mapping of anonymous memory. And the dynamic loader is gone anyway so what's the diff? I don't know if SCO kernel usied vm, but the kernel is gone. init will be attempting to complain about the absence of /etc/inittab but it can't open /dev/console. Can init even reap zombies in this state? Beats me.

When you lose all of /dev, all or /etc, and the kernel in one fell swope, it's hard to say which is the worst. If you want my guess I'll pick the loss of /dev.

But more to the point, I've seen dozens of disasters of lessor magnitude. Sometimes I've tried to recover usually by mounting the devastated disk on another system or by booting a install or a recovery medium. I succeeded about 25% of the time and only a few of those was a win. I was usually picking up the pieces for days. In contrast, it only takes a few hours to reload the os and users get a stable system much faster. So unless I'm completely convinced that I can deliver a correct running system in less than an hour or so I will always go for the re-install. I know it will work and I know how long it will take.