The wikipedia article on bang bang control (aka hysteresis control) states that one of the practical issues with bang bang control is an oscillating error signal around the desired setpoint :
First, depending on the width of the hysteresis gap and inertia in the process, there will be an oscillating error signal around the desired set point value (e.g., temperature), often saw-tooth shaped. Room temperature may become uncomfortable just before the next switch 'ON' event.
I cannot find any papers in the control theory literature that describe this issue. I'm looking for one or more published research papers that describe the problem.