The OSEMs cannot be used for coil balancing above ~10 Hz. The main coupling path from OSEM drive to sensor is not through the mirror motion, but instead direct electrical coupling of the drive wires to the sensing wires.
I put this in the elog every ~1-2 years since people keep trying it, but it keeps coming back like a zombie.
Better to use the MC angular sensors for L2A decoupling. Not perfect, but better than OSEMs. For the TMs we can use the OLs.