I imagine it will be because CV is easy to calculate, easy to conceptualise and used quite widely (so easy to compare between studies). It tends to work poorly as a descriptor of variability when mean values are very small (i.e., close to zero). We have used the median absolute deviation (MAD) in some of our studies instead, as this avoids some problems with CV that are caused by using standard deviation. Here is an example where MAD was used in a gait study: Hanley, B., & Tucker, C. B. (2018). Gait variability and symmetry remain consistent during high-intensity 10,000 m treadmill running. Journal of Biomechanics, 79, 129-134.
Starting with the second part of your answer, I agree that MAD is much more reasonable than SD, simply because that under normal distribution these should be the same, but when the underlying distribution deviates from normal distribution the MAD is a more robust estimate of the variability. Like you, I also started using it along with-, or instead of the SD (see Koren et al. 2022).
As for your answer to my question, what you are saying is that people do what other people do. This is sound when there is a good reason for this practice to begin with. The only reason I could think of is Weber’s law. According to Weber’s law, the ability to perceive a change in a signal is proportional to its magnitude. From a motor-control perspective, under a certain walking velocity there might be an optimal spatiotemporal solution. Errors in the execution of this solution are perceivable only when the errors are large enough. Large enough, according to Weber’s law, is proportional to the magnitude of the step’s metrics. So, “larger” steps will be accompanied by larger errors, simply because the system is unable to perceive them, and thus to correct. The problem with this possibility is that steps become shorter in duration, but longer in length as velocity increases, but the variance of both decreases as velocity increases.
Am I missing something? Is there a “good” reason for this practice?
Koren, Yogev, et al. "Does cognitive loading interfere with walking control?." Gait & Posture 96 (2022): 185-189.