We talk of 'dead metaphors', but we often don't reflect on what makes them die. I suspect we usually think they all die a natural death of old age. But I think at least occasionally they are murdered; a group of people seize the metaphor and stab it many times until it starts to die. Reading José Ramón Torices's "Understanding Dogwhistles Politics" (PDF) brought this to mind. The paper is actually quite good, but it shows, I think, that 'dogwhistle' is in the process of being murdered.
A dog whistle, of course, is used to train and command a dog; it is a high-pitched whistle that humans (usually) cannot hear but dogs can. They are usually just over 20 kHz to somewhere near 45 kHz. Dogs can hear the whole range (cats can hear it even better), but human hearing even optimally tops out at around 20 kHz; children with good hearing may be just able to hear an unusually low-pitched dog whistle, but in general the most any human can hear is a slight whiff of air if they are close. The exact way in which it became applied to politics is difficult to trace, but the usually accepted explanation is that it began in polling circles, to indicate the phenomenon of the respondent hearing something in a question that the pollster did not. From there it got reworked by people in political campaigns to characterize their opponents' rhetoric as deliberately being structured to rally specific audiences who could hear an implication while not alienating audiences that would not hear it. This is a phenomenon that happens. It requires an actual pre-existing jargon with specific implications. A very obvious example would be a Communist talking to a mixed audience in generic terms about fighting fascism; 'fascism' is jargon in Communist circles, and includes things like American-style liberalism, whereas for most people -- normies, as we might say on the internet -- fascism simply doesn't have the Communist meaning, and so they, unless they for some reason become fluent in Communist or the Communist accidentally oversteps so that people start getting suspicious, will interpret it as just meaning what they would ordinarily mean. Another, messier but well established and commonly given, example is that of using 'state's rights' in the American Deep South; for historical reasons, it has associations there that it wouldn't necessarily have elsewhere -- most people would take it to mean what it says, but in some Southern communities due to the pre-existing jargon a politician could use it to signal that they might be simpatico with people who try to subvert civil rights legislation, without doing so in a way that would cause problems with other communities, in which this pre-existing set of associations doesn't exist and people just assume that the politician is affirming his adherence to the principles of federalism.
It should be pretty clear from this (1) that this is a phenomenon that can and does sometimes occur and (2) that it requires a very specific set of conditions. But of course, the potential value of the metaphor for simultaneously implying that your political opponent is sneaky and manipulative while insulting the people he is trying (and presumably succeeding) to dogwhistle as trained dogs was too good to pass up. (Another reason for its being tempting is that if you can get it to stick, you can treat anything your opponent says, whatever they may say, as a really a secret code for evil things they did not say.) So, while it has had ups and downs in popularity, political commentators of all stripes have tried to characterize all sorts of political rhetoric as dogwhistles. Such abuses don't murder the metaphor, because the whole point is to keep the metaphor in place; it's just being deliberately applied inappropriately for rhetorical purposes.
The murdering comes about when academics get a hold of it, which has largely happened in the past decade, when philosophers of language decided that it would be useful to have an analysis of it. (And perhaps, given some of the things they say about it, they saw it as striking a blow for justice; philosophers have the quirk that they often think that if they have analyzed some terms they have defended the oppressed.) Unfortunately, many of the analyses that have become popular among academics stretch the term so far that it is clear that it is now covering entirely different phenomena from what the metaphor originally, and very usefully, and quite neatly picked out, and the differences are differences that do active harm to the metaphor. For instance, it's become common to talk of dogwhistles that are covert to the people who are supposed to hear them and dogwhistles that are communicated unintentionally. Both of these make nonsense of the metaphor; they are just two different ways to shoot it in the face. One might as well start talking about instances of bravely and curiously sticking one's head in the sand or of music making one's heart swell small; it's the sort of violence to a term that you'd do for a joke, not for a serious analysis of the metaphor. Obviously the salient features of the metaphor are that non-target audiences can't hear it and that you can use it rally the target audience that does. If you went around talking about a dog whistle that dogs cannot hear or that isn't deliberately used to communicate with dogs, you are obviously stretching the term -- more in the former case than the latter, but the primary use of a dog whistle is to train or command the dogs who can hear it, so if either of these are missing, then you probably need to explain yourself. It doesn't get less stretched when we do this to the metaphor; in fact, the problem gets worse, since an actual dog whistle is a multi-faceted thing, but the metaphor was applied in politics specifically on those two structural features of what a dog whistle is.
Some of the discussion that results is interesting, and does identify genuine rhetorical phenomena. And one could argue that murdering the metaphor is worth some of the resulting analysis. (For my part, I think much of the discussion is too speculative to make this argument with what has been done so far, but we could reach a point where it would be quite a forceful argument.) Metaphors, unlike human beings, have no right to life. But there is potential confusion in not seeing that you are in fact murdering the metaphor, and that the new phenomena you are identifying, the 'dogwhistles' the dogs cannot hear, the 'dogwhistles' that command without the dogwhistler commanding, are new and not the phenomena with which you started. It also raises classification worries. Are you in fact classifying all of these things properly by lumping them together when they have such clearly identifiable differences and can't be salient for the same reasons? And, of course, it raises worries about how you keep honest with an approach in which there is a category by which you can attribute meaning (usually very negative meaning) to someone that is (1) not what they explicitly said; (2) not the way they were taken by their audience; and (3) not what they intended to say. This is a true entry into esotericism. In this very paragraph, I used the phrase 'right to life'. 'Right to life' is a phrase associated with pro-life movement. One can well imagine someone coming along and saying, "Ah-ha! This is a dogwhistle! We all now know what Brandon is really saying when he talks about how analyses of dogwhistles don't respect the metaphor, and it's entirely about abortion." Wink, wink, nudge, nudge. How does one establish that this is not true, if I can't protest that I'm not talking about abortion, and no one can testify that they didn't understand me to be talking about abortion, and I can't appeal to the fact that 'right to life' is used in lots of contexts associated with murder because of course as a dogwhistle it has to be a secret meaning? Apparently it gets noticed by an infallible extrasensory perception. Such a thing makes one suspicious that we are actually looking at a different phenomenon that is being misclassified by being forced together with things more appropriately called 'dogwhistles'.
A further problem, quite closely related to this esotericism problem, is that it becomes very difficult to distinguish 'dogwhistles' from a lot of ordinary rhetorical figures of speech. A good example comes up in Torices. He uses, as an example of a dogwhistle, a comment from a speech by George W. Bush:
For so many in our country, the homeless, and the fatherless, the addicted—the need is great.
Yet there is power—wonder-working power—in the goodness, and idealism, and faith of the
American people.
As Torices notes, "wonder-working power" is a phrase that comes from a hymn; you wouldn't know it, unless you were familiar with the hymn, so he classifies it as a dogwhistle. If this is true, dogwhistles would end up being one of the most common things in politics; everything becomes a dogwhistle. It's a common (and often explicitly acknowledged) feature of political speechwriting that to appeal to an audience, you try to match the language of the audience in a register appropriate to the emotion you want to convey. If you are a speechwriter, you want your candidate to sound like he is one of the people -- and since there are lots of distinct peoples, you pick the ones that you need to rally, or that you think you can rally. You see it all the time. Bill Clinton famously used the phrase "new covenant" a lot, also a religious phrase, and one which also ties to old American heritage traditions. Al Gore was advised by his political consultants to start using phrases like "God's green earth" to increase his appeal to religious voters. Politicians have also liked terms like 'crusade' and 'faith', for similar reasons. But it's not just religion; it's everything. Politicians have tried to use gun metaphors, sometimes very implausibly, to appeal to the Second Amendment crowd, they have borrowed terms from ecology to appeal to environmentalists, they have made fools of themselves with accents and folksy local expressions, all for the same goal. At this point it becomes clear that by 'dogwhistle', you really only mean 'allusion', or, perhaps more specifically, 'allusion for the rhetorical purpose of increasing appeal with an audience'. Now, allusions are a very interesting linguistic phenomenon, and in fact Torices says a number of very interesting things with interesting implications for rhetorical allusions in general. (That we are in fact talking about allusion is clear from the fact that so much of what Torices says generalizes so nicely to allusion more broadly.) But it doesn't change the fact that we're murdering 'dogwhistle' to talk about rhetorical allusion, when we could just be talking about rhetorical allusion.