I ran the following snippet at work today and killed my CentOS Linux box.
eval 'my $' . ('_' x ~0 >> 2);
# kernel: journal_get_undo_access: No memory for committed data
VMWare was trying to write to disk while perl wanted all the ram in the universe and this blocked the disk write journal from being flushed. My disk was now read-only so I could only reboot.
On rebooting the disk was in an inconsistent state and required a fsck. It also required a root password to run the fsck so I had to call over the help desk to type that in for me.