Automatically isn't a very good idea. It would require you to insert the 'watchdog' after modifying anything, then remove it if everything is working OK. Forget to insert it and a error will force you to pull the drive; forget to remove it and a power failure a few weeks from now will undo the changes. OK, some really clever coding will ensure the undo occurs only if the two reboots occur within 2 minutes.
I prefer a second approach: Use the crypto command to set up a known password. Duplicate a good root file system to the alternate partition, modifying fstab so the proper partition gets mounted in each case. If you encounter a problem, hitting a key on your serial terminal will bring up the prom menu, and you can boot to the alternate (unmodified) partition. The chief disadvantage is you must have a serial connection.
Some of the elders (AlphaWolf, alldeadhomiez) have mentioned inserting a delay in rc.sysinit. If you hit a key in that time it will boot to a minimal system. You must have a serial connection or load network drivers VERY early.
PlainBill
There's a difference between needing help, and just being plain ole' lazy.
"You cannot teach a man anything. You can only help him find it for himself." Galileo Galilei (1564-1642)
HR20-700 with 2 TB, HR22-100, HR22-100, HR22-100, HR23-100 all running 0x5cd and networked.