Hardcore reboot

Ever got your remote system in such a spin that all the reboot commands would not work (init 6, shutdown -r now, reboot, sysrq...)
In other words: completely stuck!
If you still have the ability to execute commands and access to /proc, this did the trick for me once:

dd if=/dev/zero of=/proc/kmem

This is very dangerous to use and could do all sorts of damage... You have been warned!

busybox rules

I once accidentally removed the real /lib/libc* when building/deleting CHROOTs... ouch!
Needless to say, it becomes impossible to launch any new commands as everything on the system is likely to be linked with the C library. Therefore it is impossible to login, impossible to use tar or cp to restore the file...
As this was a remote system and I was thousands of miles away from the data-center, I was really desparate to find a solution that did not involve paying for remote hands.
After establishing a security cordon around the router that was providing my one and only remaining ssh connection to the live system, I had to move quickly, this link was not going to stay up forever...
bash command completion allowed me to browse the filesystem by repeatedly hitting [tab] in all the directories I was interested in.
Eventually, I remembered that one of my test user accounts had a static build of busybox, so I eventually found:

/home/testuserNN/build/busybox/_install/busybox

Then it was just a matter of running busybox tar to restore the deleted files... Hurray!

Since then, a statically linked copy of busybox has become mandatory on all the servers I break^W maintain.