cv::Tracker roi size

i am trying out cv::Tracker using tutorial_introduction_to_tracker.cpp, and comparing the prepackaged algorithns (MIL, BOOSTING, MEDIANFLOW, TLD, KCF). In all but TLD, the size of the ROI bounding box remains fixed for the duration of the video, whilst the postiion varies to more-or-less correctly track the selected object. I would expect the size to vary during a successful track as the tracked object depth postion varies. Is this expected behaviour? Or does it indicate that the tracker has not correctly identified the tracked object?