There’s a way of understanding the relevant alternatives theory of knowledge that makes infallibilism its ancestor. To get this result, play the “all-or-nothing” game with the skeptic so that the skeptic gets everything s/he wants if (and only if) s/he wins the game. Then, to rule out everything the skeptic will use to challenge any knowledge claim, one will have to have evidence that guarantees that anything incompatible with a given belief is false.
If one doesn’t wish to play this game with the skeptic, there are two weaker stances to take.
The first stance requires the same global reach against anything contrary to a given belief, but doesn’t require evidence that guarantees anything. Instead, it is enough to have evidence that shows that anything incompatible with a given belief is false, where it is denied that evidence shows something only if it guarantees it.
The second stance limits the class of contraries that one needs to rule out in order to have knowledge. The ruling out in question could be clarified either in terms of guaranteeing that the none of these contraries is true, or merely showing that none is true. What distinguishes this second stance is an insistence that the class of relevant alternatives need not encompass the entire range of contraries of a given belief.
In these moves from infallibilism to the modern version of the relevant alternatives theory, the central forces seem to be the abandonment of closure and the desire to preserve something of the common sense rejection of skepticism. That makes Goldman’s “Discrimination and Perceptual Knowledge” central to this history, and perhaps the first instance of the modern relevant alternatives theory (the version that refuses to count all contraries as relevant, and perhaps doesn’t require evidence that shows a contrary to be false to be evidence that guarantees that this contrary is false). But perhaps others can find precursors of this modern RAT elsewhere?