Suppose that the Humean is right that cause is merely regular conjunction. We never observe B without observing A, so we say that A causes B. However, why the ordering of A and B? If there is no objective A-theoretic fact about tense, then if causation is merely constant conjunction, obviously it can be picked out as easily working backwards through time as it can forwards. We observe A, and we observe B subsequently, and so we adopt the convention of saying that A causes B. However, B can just as easily be said to be the cause of A as A can of B. Perhaps the Humean will say that ‘A causes B’ means the very same thing as that ‘B causes A’ and that we adopt the former because it’s more useful (helps us anticipate, make predictions, that sort of thing). However, ‘A causes B’ may mean something different than ‘B causes A’ even for the Humean. For example, perhaps one never observes B without observing A, but sometimes (rarely) observes A without (subsequently) observing B. This would lead a Humean to say that A causes B, but B doesn’t cause A. I’m thinking of examples here like the quantum vacuum ‘causing’ particles to fluctuate into and out of extension. So then, the ordering of the cause matters. If the ordering of the cause matters then the Humean account (that causation is nothing but constant conjunction) is insufficient.
Note that if the ordering doesn’t matter events can obviously be causally related to things with which they are simultaneous, though the Humean might object that if they are always observed together then they can’t be empirically distinguished, but we can simply stipulate that things like quantum events provide a clear counter-example to that response.
Perhaps the Humean will argue that the notion of ‘causation’ here is too strict to allow for probabilistic conjunction (conjunction ‘most of the time’). However, that would mean that the Humean notion of causation is not only metaphysically problematic, but scientifically problematic as well; scientists cannot make use of it without either 1) making universal generalizations in science, or 2) giving up on making causal predictions for things like quantum mechanical events. Whether science involves universal generalizations may be a matter of the philosophy of science, but most Naturalists don’t construe science that way, since that’s thought to be a hangover from science’s theological infancy.
So, Humeans, it seems, should be able to accept that A and B cause each other even if A is sometimes observed without B, and B is sometimes observed without A, so long as A and B are often observed together.