OK. The particles are produced as a pair. We have been calling them anti-corelated. I mean they are opposite in some way; now presuming opposite gyroscopic rotation. They have an affinity with the rotation they were produced with that is sensitive to magnetic field orientation. The relation between rotation it was produced with and magnetic field orientation is maintained -so if the field direction is altered the particle adjusts to it, If the field is inverted the particle inverts maintaining their relationship.
Let call the particle pairs twins, and because they are opposites, twin/anti-twin pairs. (T/At). We do not know which is the twin and which the anti, the abbreviation applies to both possibilities.[To try and more clearly differentiate particles and output bit results. ]
0 degrees difference in apparatus angle. North magnet top. T/At pair.
As they were produced as opposites they have different preference of orientation in the same direction of field; such as north magnet top. They will each adjust to the field they encounter. One of the many possible adjustments that could happen is both particles turn head over heels (or heels over head) to achieve their preferred relation with the field. [Not knowing if the anti parallel axes of rotation orientation are the same for every pair or random. My guess is random.] The other extreme is they are already perfectly parallel and anti-parallel to the field and do not need to adjust their orientation. In all cases the twin/anti-twin relation is maintained. Applying the right hand rule, (N)up: (S)down or (N)down: (S)up
180 LHS N top RHS S top T/At pair
LHS same as for 0 degrees. Lets assume for ease that no adjustment is needed and (as we have to chose one of the options for discussion) The LHS particle has antiCl. Turn (rotation) giving by Right hand rule (N) up flow.
The partner finds the opposite field orientation S top. Reminder; With N top this particle, the opposite, with Cl. Rotation, would be (S) down So it maintains its preferred relation with field by inverting. Now by RH rule it is now turning antiCl. Flow (S) up. The bits are matching, up. But although the bits are deemed identical, the magnetic polarity involved in production is different.