Once I wrote a paper (two, in fact) on even Galois representations. The second paper in particular proved what I thought was a fairly definitive result ruling out the existence of a wide class of even de Rham representations with distinct Hodge-Tate weights. It turns out that almost nobody seems to cite these results, probably because they aren’t particularly useful — at least in any obvious sense. On the other hand, almost everyone who does cite the paper seems to cite it for a specific proposition (3.2) which is an easy consequence of the results of Moret-Bailly. The proposition, more or less, is a potential inverse Galois problem with (any finite collection) of local conditions. The main application of such a proposition (both in my paper and in papers which cite it) is that, given a local mod-p representation which looks like it could come (say) from the localization of a global representation associated to an automorphic form, the proposition often allows one to produce such a form at the cost of making a finite totally real extension in which p splits completely. This suffices for many purposes.
It turns out, however, that the lemma (pretty much in an equivalent form by an equivalent argument) was already proved by Moret-Bailly himself in this paper. This means that if you cite my paper for this particular lemma, you should definitely cite the paper of Moret-Bailly. Of course, if you are also applying it in a context similar to my paper (say in order to construct automorphic forms with certain local properties), you should certainly feel free to continue to cite my paper as well.