I can't figure out how the OP proposal differs from Charles Alexander's. Low frequency (long wavelength) waves travel around the tube at the speed of sound. By traveling at the same speed, and correctly choosing a point along the wavefront, one can take advantage of significantly reduced apparent drag. It's implicit in Alexander's suggestion that either (a) the total length of the loop is an integer multiple of the w…
it's slightly less wrong, as far as i can see - http://www.acooke.org/cute/HyperloopW0.html
http://www.acoustics.org/press/151st/WaveVectorDemosmall.gif
In other words: decompose the standing wave into its forward and backward components and just keep the forward one. This is not the same thing as the air having a net velocity, so you don't bleed all your energy into friction with the walls. (You'll still lose some energy as the sound wave is naturally damped by friction due to the residual motion of the air molecules, but this might not be a large effect.)