Research in training athletes revealed long ago just how critical it is that learners are able focus on only "one task at a time," so to speak. For the figure skater, for example, that means that all of the components and skills of the performance routine must be developed to a level where there will be no conscious attention required to them during the "show." The same goes for learning and training periods: no "fractured" or partitioned attention allowed.
The parallel to pronunciation teaching is apt: learning a new sound or sound process is also by its very nature a multiple-modality problem which should be a thoroughly "whole body" experience, involving all modalities simultaneously. If it isn't, there is a very good chance that, just as in figure skating, it risks not engaging enough of the mind and brain for the target to be learned efficiently--if at all. Krashen (1972) and others were right that "monitoring" can be quite destructive to fluency and learning, but it goes considerably further.
So much of contemporary pronunciation instruction is at best "dis-integrated," and even more likely to be relatively dis-embodied, as well. That is especially the case with methods that over-rely on listening, explanation, uni-sensory repetition and non-systematic "learning-in-real communication" -- accompanied by the inevitable, random, multi-tasking and compromised, temporary, partial attention.
That "figures," doesn't it? They are just out of touch . . .