Christianity is heading for a full allegorization

[Angrboda]

How does any of that matter?  Every summer drownings correlate with ice cream trucks being in the neighborhood.  It being local or not is irrelevant because the speed of light need not be violated by the causative effect of ice cream trucks.  QM postulates statistical effects so the idea that ice cream trucks correlates with drownings need not occur at the individual level.  Additionally, that assumes that 100% of drownings are caused by ice cream trucks and that there is no delay, fixed or random, in the mechanism.  It hypothesizes exactly what we see, a rise in frequency of nearby ice cream trucks correlating to a rise in drownings.  The fact is things correlate with other things regardless of how near or far, aggregate or individual, or any of this other crap.  You're trying to add other factors on top of correlation as being necessary to establish causation, but your prior argument doesn't allow for that.  Your Humean skepticism has led you to a dead-end in which you can't rule out anything as a cause of anything else.  I just burped.  Somewhere in the world, somebody fell off a building.  And according to you, I caused that, because my burp correlated with them falling off a building.

We have two hypotheses:

1. Ice cream trucks cause drownings.

2. Both ice cream trucks and drownings occur mostly in summer.

So what we need is an observation that will distinguish between these two hypotheses.

The most obvious one would be to rent a bunch of ice cream trucks in the winter and see if drownings increase then as well. Even better, do this at various different times of the year and in various locations. Then see if the correlation persists.

You are right, the correlation is primarily observational.  But when there is more than one active hypothesis (and there almost always is), the key is to find some setup where the two hypotheses give different predictions and see which one is wrong.

This is yet another reason why a single experiment is not enough to overthrow a theory. The experiment needs to be conducted in a variety of situations to explore when the observed correlation (or lack) is there.

I think the basic problem is the expectation that science will give a 'mechanism' for all 'causes'. And that is simply false. In fact, the whole idea of 'mechanism' assumes a metaphysics that is very likely to be wrong.

How are you ruling out the possibility that cold temperatures can cancel out the ice cream truck effect? You have a methodological flaw in that your experiment can't distinguish actual falsification from confounding factors. All you can say is that it is unknown why ice cream trucks do not cause drownings in winter, not that the hypothesis that ice cream trucks cause drownings is false. This is because your test, and any test you devise, depends upon the absence of certain observations, and that absence could have many explanations. Science depends upon observations, not inferences based on their absence. As noted earlier, basing a conclusion on an absence of something is basically an argument from ignorance and is invalid. Cold fusion has never been observed. Does this mean that cold fusion does not occur? No, it does not. I think the basic problem is there is no prediction implicit in a correlation, only that under certain conditions, the two phenomena co-occur. This does not lead to any type of prediction as the correlation doesn't imply that the two will co-occur under any conditions, so any changing of the conditions will leave you, not with a falsification, but simply an interesting mystery. Explanations, however, possess the property that certain inferences naturally follow from the explanation and thus predictions can be made. But without an explanation, you're simply left not knowing which correlations are meaningful and why, and the inability to produce predictions from the correlation means that you can't falsify the theory. If all theories are nothing more than correlations, then no predictions can be made from them, and none can be falsified. That would mean that none of them are scientific.