Jonathan Blow (of Braid fame) has an insightful response to Stephen Totilo's comments lamenting the lack of gaming landmarks: the status of "landmark" shouldn't be tied to representations of fantastical architecture or a particular visual look, rather to what happens in those spaces. He talks about landmarks of "conceptual space": I started having bad flashbacks to slogging through Benedict Anderson's classic Imagined Communities at this point, but Blow has some interesting points and examples (he points to Counter-strike and Team Fortress maps that may change their look from incarnation to incarnation, but retain a sense of place thanks to the history of gameplay within those spaces, no matter what form their visual trappings take):
... After sleeping on the question for a few days, this occurred to me: if we judge landmarks by their visual impressions, we tend to ignore what games are about, a large part of which is interactivity, and the player's understanding of the way things work within that game world. If locales are really going to be game landmarks, rather than fanciful imitations of real-world places that you could experience as well in non-game media, then the impression they leave needs to happen through gameplay; they need to be memorable because of the things they encourage to happen within them, not (just) because of the way they look.
So what makes a gaming landmark for you? I had a hard time thinking back to any sort of iconic structures, but I have plenty of games that have created such a strong sense of space that they would be included in my personal 'gaming landmarks.'
Landmarks, of sorts. [Braid]