For example, say you wanted to set up computers for public use at a school or library. Which way is it better to set the option? If it were a default/option pair, it would obviously be better to set it to the default, since people would be more familiar with it. If, however, it were just two equally valid options, which one would people be more likely to be familiar with? Multiply this by every possible [add-on on/add-on off] pair, and you have, perhaps, 1000 or more default browser configurations, any of which are equally likely to run into in the wild!
Anyway, on your first point, I think users need to be a little more sensitive to the fact that, if they don't allow themselves to be pushed out of their comfort zone every once in a while, their experience using their software (or anything for that matter) will never improve. If a feature worked crappily, one would never know it until they adapt to the "new way" enough to see how bad it was before. Until then, their comfort with the old way makes them much too partial to be a good judge.
That's the thing actually - supporting the old behavior in case of FF3 wunderbar is trivial as it's just a subset of the new behavior. Just restrict latter as this:
(a) search for a match only among the URLs
(b) match at the head of the URLs; skip "www." when matching if it's present
I'm not an FF dev, but this should hardly take more than a couple of hours of work, esp. if one is familiar with the code.
I tend to agree though. I'm still being driven nuts that Google's favicon has changed. I think companies need to be sensitive to all the momentum that people bring to their products.
But then you have to balance that with the overhead of supporting an "old" feature for the sake of old users.
This might have been one of those things that might have worked better as an add-on. I think it might have been kinda slick when you first fire up FF3 to be presented with a choice of basic add-ons to play with, like "Smart Bar".