(1) A female graduate student is constantly addressed by her male adviser through terms like "honey" or "sweetheart".The first one is gendered on face, the second is not. This matters.
(2) A female graduate student feels constantly belittled and patronized by her male adviser. He always seems to talk down to her and does not view her as a worthy member of the intellectual community.
A skeptic of microaggressions, responding the former scenario, might concede that there is a gendered aspect to the comments but argue that any "offense" was probably unintended and in any event is marginal. "If this is the worst you're experiencing in terms of sexism," the argument goes, "count yourself lucky." This generally overlooks the pervasiveness of the statements. Pretty much everyone agrees that any singular instance of an activity termed a "microaggression" is not itself all that harmful. The problem is in the pattern: as Prof. Rini puts it "over time a pattern of microaggression can cause macro harm by continuously reminding members of marginalized groups of their precarious position." If someone made a weird joke about how my being left-handed marked me as the devil's servant, maybe I'd find it funny or maybe I wouldn't (depends on how good the joke was), but seeing as it would be the only time in the past decade someone had made reference to by leftiness in that way I'd probably shrug it off either way. If I lived in a different society where such "jokes" were routine and there were other broader barriers to lefty-equality and social sentiments promoting discrimination or exclusion of lefties, I'd feel different.
In any event, what matters in this case is that the cards are generally on the table. The operative question is how commonplace the slights are, and that is something the woman in question has directly experienced. I don't want to say there is no room for self-doubt -- we can still ask ourselves "is X number of gender-demeaning statements sufficient such that I am warranted to feel aggrieved" (contrary to popular belief, most women are not like Venus Fly Traps eagerly awaiting a stray call of "honey" to cross their path so they can experience the joy of crying victim). But most of the critical facts are pretty straightforward; indeed, I think it is fair to say that the microaggression critic here is in many cases simply unaware of the raw numbers and wrongfully views each microaggression in isolation.
Now consider the second scenario. Here the skeptic's response would be quite different, asking how we know that the professor's conduct relates to the student's gender? Maybe he's just a general asshole (hardly out of the question). Or maybe the student really isn't that bright (it happens, sometimes). Or maybe there is an other non-gendered factor causing his disdain. Moreover, it's not like male students don't also sometimes feel like their advisers don't respect them or speak to them as equals. Who's to say this instance is gendered?
Note that the problem here is in some ways the opposite of what is going on in the first case. Most people would agree that being derided and degraded by one's adviser (at least without good cause) is itself a pretty miserable experience -- even as a one-off. The dispute is over whether this wrong can be attributed to gender. By contrast, in the first set of microaggressions, the gendered component isn't really disputed, the argument is about whether the actions are pervasive enough so that one can characterize them as a wrong (or at least, a wrong not so trivial so as not to be worth talking about).
In a laboratory, answering the "how do we know" question also would generally be a matter of aggregation: when trying to unpack the roots of hostile conduct that could have both identity- and non-identity related causes, one thing we look for is the variance in how diverse groups are treated. If 75% of women in graduate programs are patronized but only 35% of men are, the evidence of sexism lies in the gap.
But people are not laboratories. Everybody knows, of course, that everybody sometimes receives bad treatment from their superiors. The woman in question may suspect, based on her experience, that it happens to her or her female peers more often than it does to men -- but this is a much more indirect and tenuous observation than would be the number of times she is called "sweetheart". Moreover, even under ideal laboratory settings the fact of the variance is evidence of discriminatory treatment existing, but it provides no hint as to which particular cases are and are not gendered (indeed, it strongly implies that some are not gendered). A woman trying to discern whether her particular experience derives from sexism or her own inadequacies or her adviser being a general jerk or something else will no doubt have to rely on much more subjective elements of her own judgment in making her evaluation.
This would all be tough enough if we respected the judgment of women when making these calls. But we don't, and this is where I think an epistemic injustice comes into play, and, more broadly, why I think the second class of microaggressions poses a particular peril. Epistemic injustice is probably the only reason why the first type of microaggression receives the doubt that it does ("Oh come on, you couldn't be cat-called that often. You're exaggerating!"), but it really rears its head with a vengeance in the latter case. The opinions of women or other marginalized groups -- in general but especially in this context -- are routinely dismissed as unreliable, self-centered, narcissistic, over-sensitive, or just plain nuts. In that world, a type of microaggression which can only be identified at the case-level by reference to subjective judgment is going to slam into the broader social belief that these judgments can't be relied upon. There isn't going to be a smoking gun in these cases, and the general rule is that without said gun marginalized persons are going to be ignored. And this message, trumpeted loud and clear, can knock one's own confidence in one's own appraisals of the surrounding world -- "am I just an over-sensitive whiner complaining about the same slings and arrows every person faces regardless of gender?" (Kate Abramson has a stellar paper exploring this topic through the lens of "gaslighting").
All of this leads to one more oddity I want to raise. Sometimes, I'll be engaging with a person or group that seems to me to be anti-Semitic, but hasn't said or done anything explicit yet -- my instincts are based on suspicions of differential treatment or lack of respect, but I can't outright falsify the possibility that Jewishness really has nothing to do with their positions. And then sometimes the person or group will say something that seems to tear the mask off -- positively quoting from Nazi propaganda or urging schools to expel Jews -- making the anti-Semitism clear and unambiguous. And even though this is a move to a more serious wrong on their part -- it converts a microaggression into a macro- one -- there is a part of me that often breathes a sigh of relief. There -- now that I have that smoking gun, people will believe me (I alluded to this issue in my Innocent until Proven Nazi post). And there is an even smaller part of me that breathes a sigh of relief and says See, you're not crazy after all. Even someone like me, who writes and talks constantly about the injustice of the default assumption that Jews and other minority groups either lie or are deluded when commenting on their own oppression, still has partially internalized the omnipresent social drumbeat that my instincts on anti-Semitism are systematically unreliable.
In short, there is a significant class of microaggressions which are genuinely ambiguous at the case-level. But the problem is not overzealous women, or Jews, or people of color, who seek to trample all over this ambiguity in a race to claim victimhood. What happens more often is that these people will perceive a problem, but be aware of the ambiguity, and know that they can't muster incontestable evidence to prove the problem, and are well aware that their subjective judgment on the matter won't be valued, and so are completely boxed off from any socially-sanctioned mechanism for talking about the perceived wrong or even fortifying their own epistemic self-regard. They are left in a limbo state, trapped with the nagging belief that maybe what everybody says about them is right, maybe they really are just making it all up. And that danger -- compounded by the fact that we lack socially-sanctioned mechanisms for talking about perceived wrongs in cases where there isn't smoking gun evidence -- is a real one and I think in many ways the most serious problem microaggressions pose.