Is there any compelling reason to update the firmware if everything is working OK to begin with?
Who said it was working OK to begin with?
I've been chasing a random wireless client disconnect problem with MT gear for over a year. Also, unless you've got a good 40 dB SNR, AP+WDS isn't exactly what I'd call reliable. 3.x (3.2 especially) is considerably more reliable on the wireless end of things, its just the upgrade procedure thats broken (IMHO) at this point.
The only ways I can see MT correcting the routerboot problem:
add a checkpoint to the installation script to ensure the routerboot. You could handle it two ways after this:
A> just flash it and move on with life
B> stop the rest of the installation from occurring (similar to loading the wrong architecture to the routerboard)
Either way, I see this situation easily correctable. It'd also help if MT didn't 'misrepresent' very experimental builds as 'stable releases'. I don't consider 3.x 'release quality' by any means, and I can't understand why MT rushed the release other than to be able to finally ship their 3.x-only hardware without telling their customers all they have available is 'release candidate' firmware.
As to answer the original question: I upgraded the routerboot because the unit would no longer recognize its installed SR9 after the 3.x upgrade.