The tubes would have to be exactly the same length. Mercury delay lines are, I think, fairly well known. Any difference in the tubes would result in the signal to each speaker arriving out of phase with the other. But then what about the small but surely audible difference in timing between the axis and the radius of the mercury columns...?