Mouseferatu
Hero
I'm going to go against conventional wisdom here and say that we have not reached the saturation point on D&D/D20 settings*. In fact, I'd suggest there's no such thing--if settings are done "correctly."
*(I'm referring here to settings that are intended for use with Dungeons and Dragons as written. I've no intention of discussing entirely OGL settings like Conan here; that's a whole different kettle of wax.
)
Everything easy has been done, as have a number of things that aren't easy. I believe a setting can still be successful, but it's got a lot of hard work ahead of it.
A new setting must be sufficiently different and/or interesting to attract attention. Another "generic fantasy world" in the vein of Greyhawk/Forgotten Realms/Krynn/Mystara is simply not going to cut it. A setting has to have a hook, and "Like X but better" doesn't cut the mustard anymore.
Now, here's the hard part. It's easy to be different. Being different isn't enough. Despite being a different, a new setting, to be truly successful, must cleave to the core rules as closely as possible. New settings can add options and rules all they want, but as soon as you start removing fundamental aspects of the game, you've sliced huge chunks off your potential user base.** If you want to remove or change core aspects, you've got to do so in a way that can be ignored; make them optional, or at least provide optional rules for counteracting those changes.
**(Some very successful settings have violated this rule. I maintain that this is, at least in part, because they came out when they did, and that any attempt to do so today would fail. I'd also point out that even successful settings that do this still lost at least some potential customers. I love the look and feel of Midnight, but my group would never play it. When we pick up a D&D setting, we want to play D&D. Midnight eliminates or changes too many of the core classes and rules to appeal to some people who want a different setting but not a different rule set.)
A new setting has to be widely usable. A "drag and drop" setting like Freeport is one way to do it: Something that can be placed in an ongoing campaign. This isn't the only way, though. If a setting provides sufficient flavor and usable material, it can appeal to people who don't plan to use it whole cloth.
A new setting has to be self-contained. While supplements are fine and dandy, it should be fully playable with a single core book--maybe two, if you really stretch it. More than that, you run into the problem Psion astutely mentioned regarding Scarred Lands: There's a large barrier to entry.
Now, notice what I didn't say: I don't believe a new setting automatically has to cover brand new ground. It certainly helps if it does. And if it covers territory that's already been covered in another setting, it has to offer a hook beyond "Like X, but better," as I said above. But that doesn't make it impossible. An Arabic setting's been done. A psionic-centric setting's been done. Multiple horror/dark fantasy settings have been done. That doesn't mean those particular wells are dry. It just means a new setting that covers the same ground has to offer something different--an approach, a perspective, mechanics, whatever--to draw attention.
A new setting that successfully accomplished all of the above is exceedingly difficult, but not, I think, impossible. And so long as a new setting can manage the above, I think there's always room for one more.
*(I'm referring here to settings that are intended for use with Dungeons and Dragons as written. I've no intention of discussing entirely OGL settings like Conan here; that's a whole different kettle of wax.

Everything easy has been done, as have a number of things that aren't easy. I believe a setting can still be successful, but it's got a lot of hard work ahead of it.
A new setting must be sufficiently different and/or interesting to attract attention. Another "generic fantasy world" in the vein of Greyhawk/Forgotten Realms/Krynn/Mystara is simply not going to cut it. A setting has to have a hook, and "Like X but better" doesn't cut the mustard anymore.
Now, here's the hard part. It's easy to be different. Being different isn't enough. Despite being a different, a new setting, to be truly successful, must cleave to the core rules as closely as possible. New settings can add options and rules all they want, but as soon as you start removing fundamental aspects of the game, you've sliced huge chunks off your potential user base.** If you want to remove or change core aspects, you've got to do so in a way that can be ignored; make them optional, or at least provide optional rules for counteracting those changes.
**(Some very successful settings have violated this rule. I maintain that this is, at least in part, because they came out when they did, and that any attempt to do so today would fail. I'd also point out that even successful settings that do this still lost at least some potential customers. I love the look and feel of Midnight, but my group would never play it. When we pick up a D&D setting, we want to play D&D. Midnight eliminates or changes too many of the core classes and rules to appeal to some people who want a different setting but not a different rule set.)
A new setting has to be widely usable. A "drag and drop" setting like Freeport is one way to do it: Something that can be placed in an ongoing campaign. This isn't the only way, though. If a setting provides sufficient flavor and usable material, it can appeal to people who don't plan to use it whole cloth.
A new setting has to be self-contained. While supplements are fine and dandy, it should be fully playable with a single core book--maybe two, if you really stretch it. More than that, you run into the problem Psion astutely mentioned regarding Scarred Lands: There's a large barrier to entry.
Now, notice what I didn't say: I don't believe a new setting automatically has to cover brand new ground. It certainly helps if it does. And if it covers territory that's already been covered in another setting, it has to offer a hook beyond "Like X, but better," as I said above. But that doesn't make it impossible. An Arabic setting's been done. A psionic-centric setting's been done. Multiple horror/dark fantasy settings have been done. That doesn't mean those particular wells are dry. It just means a new setting that covers the same ground has to offer something different--an approach, a perspective, mechanics, whatever--to draw attention.
A new setting that successfully accomplished all of the above is exceedingly difficult, but not, I think, impossible. And so long as a new setting can manage the above, I think there's always room for one more.