Looking around https://stellartech.science/en/qeq/ I found the strategy they implemented to deal with backlash interesting. Basically they 'accelerate' the movement in a given axis upon swapping directions for the corrections, eating the backlash in a much shorter time. 

I wonder if something like that could be implemented via software in Ekos for the autoguider module. Basically I imagine one could have either a few boxes to manually set the backlash for each axis and each direction, or alternatively, the system could figure out itself the backlash during the calibration sequence.