Question: What are the features and failings of the "software" brains must use to control muscles?
My hypothesis: Controlling muscles and skeletons in 3-D space requires motor maps which ideally estimate current configurations in continuous space and extrapolate them into new configurations nearby in space and time. Mis-learning caused by inadequate or bad data results in maps with unaddressable muscles (sensorimotor dark-space) and spatio-temporal discontinuities, which appear as the crackling or snapping inside joints during continuous motion. See Continuous Models.
Summary: Most muscular pain and motion limits are data problems, not tissue problems, and suffer from the same feedback traps as data does.
Refutation: Reproduce the popping, snapping, and inflexibilty of joints when cortical control and/or sensation is disabled.
(My personal experience and second-hand knowledge from conversing with credentialled bodyworkers so far agree with all these tenets)