A few months ago there was a next release with a bug in the magnetometer calibration: instead of pressing "save position" you had to press the "save" button in the lower right hand corner of the gcs.
That was the first time i had a calm, oscilation free horizon.
Probably the calibration routines changed to something worse in rc1/2?
Observation:
Setting home location gives an "gravity accelleration" of 9,81000042 in my location.
Reading out AccelState z-axis in gcs< Data Objects shows a value fluctuating between -9.87 and -10.00, around the same tempo of fluctuation as the wobble in the artificial horizon.
Meanwhile the z-axis MagState is showing smaller and slower fluctuations from 478 to 481 mGauss.
I suspect the magnetometer and other calibrations need to be looked after.