Step two of the cited argument from Sider is a step that relies upon a certain link between conceivability and possibility. It could have been split into two steps and phrased thusly: "2a) Gunky objects are conceivable 2b) (therefore) Gunky objects are possible". Possible worlds seems to play a merely heuristic role here.
There is, however, a more general and interesting question about whether possible worlds talk grounds or makes true modal claims--- the question of which is metaphysically prior you might say. The metaphysical/explanatory priority of possible worlds is something Lewis would accept, but I'm not sure how common a stance it is. It relies on a certain reduction that only the modal realist has been able to pull off (and even there you have critics claiming the reduction is unsuccessful).
The point here turns on whether you take modality as a primitive, or attempt to analyze it. There have been many attempts to analyze modality, but Phil Bricker's "Reducing Possible Worlds to Language" has convinced many (most?) that it is not possible to reduce modality in an actualist-friendly way. The problem is that a language that countenances only what is actual--- and tries to account for the possible in terms of these things ---will find itself unable to distinguish between certain "alien possibilities". These are possibilities containing individuals and/or properties which are merely possible and indistinguishable in actualist terms.
Now, you could simply accept a possibilist language and reject the Lewisian modal realist picture--- but then you haven't given an analysis. What you would have done is made the intensional modal language extensional. In fact, this was one of Lewis's original claims in favor of his Counterpart Theory. In the postscript to "Counterpart Theory and Quantified Modal Logic" in his Philosophical Papers vol. 1 Lewis, however, disavows this as a positive reason in favor of counterpart theory since he no longer sees anything wrong with intensional languages. The section of Bricker's "Concrete Possible Worlds" titled "No Primitive Modality" (starting on page 6) contains useful discussion of what is meant by "primitive modality" as well as Lewis's own attempt at reducing modality. This summary of On the Plurality of Worlds might also be helpful to you. Pages 6 and 7 of that summary also give some of Lewis's reasons not to hold modality is primitive. This Sider paper surveys a number of attempts to reduce modality along with some standard objections to these attempts.
All of this has ignored one central aspect of your question, namely, the epistemology of modality. This SEP article provides a nice overview of the area, and you can find many more articles in the bibliography there. The canonical anthology in this area is Conceivablity and Possibility. As far as I know (though this isn't my area) there is no one who holds that we know of modal truths in virtue of knowing truths about possible worlds. Virtually everyone (I think Lewis included) would hold that knowledge of modal truths is epistemologically prior to knowledge of possible worlds.