Read what lxo has written and tell me if his reasoning survives if you remove ideology. As for the rational reasons for firmware source being generally benefitial, they are obvious and mentioned in this thread. Starting with potentially improved odds of figuring out why given piece of hardware sucks in given way by whoever ends up having to debug the breakage caused by FPOS in question. Balanced by the risks of nausea from reading it, of course, but that's a separate story.
But the same odds are also improved by
* having hardware documentation that is not entirely full of crap
* having firmware not written in Object Intercal implemented in BLISS macros
* having no"! yes, sometimes the usual opcode for NOP deadlocks the branch predictor; we'll do a hardware fix someday, for now let them patch as(1)" in there,
etc. All of those are generally good things. And if some are *not* true, you'll probably have more good reasons to curse the bleeding vendor. Which doesn't mean that e.g. removal of fuloong 2f support would be an improvement (NOP breakage, making life rather interesting).
IOW, "P may make dealing with Y less painful" doesn't mean "P should be a prereq for doing anything with Y".