Kumaravadivelu identified the last "system" in language teaching, the last prevailing method where internal changes could be judged in terms of effectiveness: the structuralist "Audio-lingual" paradigm. It has (thankfully) nearly disappeared today. Its problems with generalizability were legend, but something also was lost: a common method where individual variables and techniques could be credibly assessed for effectiveness. Tucker's argument speaks clearly to our problem today.
Problem? Well, maybe it is also an opportunity for individual instructors to maintain perspective when reading research studies focusing on one variable or technique before trying it out on students--and more importantly trying to figure out whether something worked or not. ("Research" has overwhelmingly established that it is always far more difficult to learn from our successes than our failures.)
What is the solution? My guess is that a new paradigm, a more iconoclastic method--for teaching pronunciation in this case--will emerge from the chaos. What would that look like? Like ALM, it will at least initially show promise to provide a highly systematic model, a more comprehensive and complete set of tools for a wide range of learning populations and classrooms.
At the moment I can (not surprisingly) only think of one . . .