I watched this YouTube video by Sabine Hossenfelder to try and better understand superdeterminism:
(Note: Physics begins around 8:12, before that she's mostly talking about history and background information.)
At 12:11, she defines Superdeterminism as follows:
What a quantum particle does depends on what measurement will take place.
She then introduces this equation: $$ \rho(\lambda|ab)\neq \rho(\lambda) $$
She explains it as meaning that the probability distribution of the hidden variable changes, depending on the settings of the detector at the time of measurement.
Since the hidden variable exists before the time of measurement, it sounds like superdeterminism is nothing more than a retrocausal explanation, where the measurement retroactively alters the hidden variable. But I must be misinterpreting something, because retrocausality is (relativistically) equivalent to superluminal communication, and would therefore violate locality. At the beginning of the video (1:42), she explains that locality is a separate assumption from statistical independence (i.e. "no superdeterminism"), and at 17:05, she explicitly states that superdeterminism is local, and so a retrocausal explanation would seem to be ruled out.
The only other interpretation I can think of is that the hidden variable and the detector are correlated, not by some causal or physical connection, but as a brute fact, with no underlying physical mechanism connecting them. I find that interpretation difficult to accept, but I suppose it's not directly contrary to any law of physics. This is what I had believed superdeterminism to be before I watched the video, but now I'm confused.
Is superdeterminism a retrocausal explanation, or is there some other physical mechanism that could explain how this correlation arises?