I think the two are related—that is, the overarching goal is to optimize people’s ability to get help (really, one could argue the real goal is poverty reduction).
Someone’s engagement with a universe of services is colored by uncertainty about their ability to access services. So, in the case of something like SNAP, regardless of whether someone is made aware of the requirement, the requirement still exists. I don’t think omitting it entirely is the right move. Regardless of whether we plan for it, people already build mental models about any system they use. And so hiding the explanation completely might leave people to conclude that a system isn’t made for them.
But you’re right that we can’t present everything, all at once to someone. And even if we do, it doesn’t mean people are going to build the right models in their heads about a system. So my answer to your question is incremental, intermediate (and perhaps optional) explanatory models. Good design might be matter of helping a user acquire that knowledge one step at a time, if they so choose.