(1) anchoring the basic moves
(2) instructor and learner working together to integrate the moves
(3) learner "solos!"
In PHASE ONE, learner works on the felt sense of tossing (i) one ball in one hand, (ii) one ball going back and forth between hands, and then (iii) a second ball is introduced in the hand that will catch the other ball and be tossed away just before "main" ball arrives. The haptic parallel is basically anchoring the essential movements of the target sound without attempting to coordinate them. (There are rarely more than three critical parameters.)
In PHASE TWO, learner begins to combine features as the instructor/coach responds when needed in achieving accurate individual movements. Next learner and coach juggle/do the sound together. In the process, the learner's attention is directed away not only from environmental distractions but also from focus on the mechanics of each parameter, which is becoming more automatic and non-conscious. The instructor/student dance does much to enable that integration.
In PHASE THREE, learners "juggle" the new sound on their own. I have not seen a better model (or metaphor) for changing pronunciation. So should you learn to juggle first or simply "juggle" your teaching? It's probably a toss up . . .