After swapping the magnets on the beamsplitter, cf 8121, I tested the middle mass (M2) of the suspension by sending offsets to each osem, expecting to see the suspension being pushed with a positive offset, and vice versa.

Results are attached below with (1) the generic drive signal sent consecutively to each osem, and (2) the response (in um) of each osem to the drive. 

Not sure if it make sense that the lower right osem has a smaller amplitude than the others, but other than this the suspension is moving as expected.