This is a difficult epistemological argument that it took me some time to accept, but: how can you confirm the explanation is correct, if not by testing it as a hypothesis, i.e. using it to predict unknown (future) outcomes?
Put differently, if you say "X causes Y, but only in this one instance we're discussing now and not in the future" then does X really cause Y, or was it a coincidence that they happened at the same time this once? The razor leads us to conclude that what's indistinguishable from coincidence is coincidence.
----
Edit: Oh, I see what you're saying! Something like "X causes Y but knowledge about X is only available after Y is already known"?
That is possible, and it's even testable by blinding the person to Y while revealing X. Is it useful knowledge? Maybe!
Put differently, if you say "X causes Y, but only in this one instance we're discussing now and not in the future" then does X really cause Y, or was it a coincidence that they happened at the same time this once? The razor leads us to conclude that what's indistinguishable from coincidence is coincidence.
----
Edit: Oh, I see what you're saying! Something like "X causes Y but knowledge about X is only available after Y is already known"?
That is possible, and it's even testable by blinding the person to Y while revealing X. Is it useful knowledge? Maybe!