Speculative fiction is a genre rife with things. In sci-fi you've got your starships and your holodecks; in fantasy you've got your celestial swords and your infernal incantations; and in horror you've got evil artifacts and haunted graveyards. Lately, I've noticed--in both my own writing and the work of others--that this prevalance of supermundane objects creates for speculative fiction writers an omnipresent variation on the problem of Chekhov's gun: an item displayed at the start of the story plants a seed of expectation that the item's powers will come to fruition by the end.
I realized this by picking up on a bunch of false-guns in Critters critiques, and now I'm paranoically hunting my own. The sheer amount of errata that can be thrown into a sci-fi or fantasy story means that there are so many opportunities to create unintentional Chekhovian devices. By doing this, you can develop unwitting, tacit promises with the reader that send their expectations in another direction than your own, and hence ruin the story by not fulfilling the reader's desires. In my Critters work, I've bumped into a bunch of stories that let me down because I was led to believe some thing or some person was more important than it really was.
I've realizied that the only way to avoid this problem in my own writing is to monitor what kind of stuff--from minutiae to unobtainium--that I'm delivering to the reader. Anything that's repeated, anything that receives more than a sentence of attention, anything that the protagonist touches, breathes, or thinks: this will attain a degree of importance and be retained in the mind of the reader. "Sore thumbs" are the worst: powerful items that can "get'er done," and insanely unique locations, creatures, people, and collectives. It's very easy to dwell on the aspects of a futuristic city or a magical order, but--especially in short fiction--descriptions like this can't afford to be sidebars. If the city or the order isn't important to the resolution of the story's essential conflict, the amount of attention it receives must be stripped to the minimum.
I think the best thing to do is to put the meat on the bones of your story by referring deeply only to those elements you can at least resurrect, in some capacity, later on in the story; everything else should be skipped over or mentioned only in passing. This way, the expectation you create in the reader--the feeling of "I want to know more"--can be satisfied. For example, I'm writing a story at the moment where the denizens of a particular realm have a fear of taking weapons--or hostilities more generally--across bridges. This fear stems from the fact that in ancient times, terrible monsters guarded the bridges, and the fear has outlasted the monsters. In the story in question, I described these ancient monsters, and their present absence, before it had any relevance to the plot. But later on, I saved the protagonist by utilizing this mythological fear of bridges to keep a mob at bay. This also ensured that the Chekhovian device--bridges--resolved its internal tension within the story.
Worry might arise here: how will the reader understand the basic functions of a magical or futurological world if you don't take the time to describe them? Case in point: in the story I mentioned above, the bridges connect hunks of land that are simply floating in space. Maybe a reader will wonder about the physics of this world, and if I'm a good writer, I hope they'll want to know more; but I think they'll be more inclined to simply accept that it's a fantasy story, and allow themselves to be drawn into and enjoy the objects that receive greater attention--like those bridges.
And, of course, all the other random stuff I described in the first draft--the stuff that didn't have a role to play in the conflict's resolution--got chucked. Writing in this way--removing anything and everything that doesn't pertain to the matter at hand--will not only avoid making false promises to the readers; it also keeps the story tight and driven.