Dual Firmware

These methods are currently theoretical and have not been tested as of yet.

Hardware Based

NOR/Nand Piggybacking

This method involves physically soldering another flash chip ontop of the existing flash packages, soldering the legs pin for pin (piggybacking). You will lift both #CE pins and provide a switch between them to select the appropriate flash chip, of which each will have a different firmware.

Reset pin for NOR

After looking into this some more, Simply switching the #CE pin may not be sufficient as the chip is still operating and can interfere with the bus. However, it appears that whilst the #reset pin is tied low, all input/output pins on the flash are in a state of high-impedance. We should be able to simply ground this pin to disable that chip, rather than lifting the #CE pin.

Dual-Banking

This method relies on the fact that SYSCON has 2 EEPROM banks, and a "recovery mode" flag that can be set to load a recovery firmware located in the ros1 region of the flash.

By pulling the backup_mode pin low or high, you can aparently switch eeprom banks in the SYSCON EEPROM. In the second bank, you would have the recovery mode flag set, thus loading firmware from the ros1 region on flash.

Increased size NOR Flash

This method relies on entirely lifting the existing NOR flash chip and planting a 256mbit chip, you could lift Address pin 23 and have a switch to tie this low or high to switch banks. A compatable samsung chip can be found below: http://www.samsung.com/global/system/business/semiconductor/product/2007/8/7/620430ds_k8p5615uqa_rev11.pdf

This looks like it could work, as per the spansion and samsung charts, when using autoselect commands etc, it does not care about the state of pin 23. So there should not be any interference.

Limitations

Firmware hash checks

Firmware hash checks are located on SYSCON EEPROM, aparently these checks are run within Indi info manager on LV1. These compare the hashes stored in syscon with the files stored on flash. If the checks fail, the console does not boot. We could get around this by using dual-banking on SYSCON or by patching the checks out.

VFlash

Only a single version of VFlash is stored on flash in NAND consoles, and a single copy is stored at the beginning of the PS3 hard drive on NOR consoles. Because the firmware stored here doesn't match that stored on flash, you would have to reinstall the rest of firmware everytime you switch. We could possibly overcome this limitation by patching the storage manager to redirect vflash to another region of the hard disk.

Software based

Using graf_chokolo's payload

In graf_chokolo payloads, there is a payload that can be used to load an alternative lv2_kenel.self

You have to save the alternative lv2_kernel.self on flash and use the payload to make lv1 load it.

See Graf's PSGroove Payload

Quoting graf_chokolo

Guys and be careful with store_file_on_flash.c and replace_lv2.c payloads. 
With store_file_on_flash.c i’m able to store a new file on FLASH memory where CORE OS files are stored from PUP. 
If you do not know what that means then don’t play with this, it could brick your PS3, but it’s safe to use when you know what you do.
With both of those payloads i’m able to boot a patched lv2_kernel.self from FLASH without flashing PUP, i just store a second lv2_lernel.self
on FLASH, then patch System Manager in HV which is reponsible for booting GameOS and boot custom LV2 kernel from 3.41. 
You don’t need NOR flasher if something goes wrong, just reboot HV and your original lv2_kernel.self will be booted again

The same way you could boot lv2_kernel.self from dev_flash. Just patch path to lv2_kernel.self in System Manager and point it to lv2_kernel.self stored on dev_flash

Limitations

• Same as above and this could ONLY be used with a lv2_kernel.self compatible with you actual lv1.self
• You can only customize lv2_kernel.self and below

Bootloader

There is master hardware based different for every PS3. It is said that some have managed to get ahold of it. Being able to sign in a higher privilege would give us the ability to create a bootloader that would allow us to load any firmware previously patching it.

This would be the best solution, having a bootmii like bootloader with recovery options, but it is also the most farfetched.