an audio signal is represented electrically with a voltage difference between two conductors. if you reverse the polarity -- swap which conductor is positive and which is negative -- the measured voltage is essentially multiplied by -1. If you think about it as a sine wave on a graph, the whole thing is flipped across the x-axis.
if you were to take a signal, reverse its polarity, and add it to the original, the entire thing would cancel out and go to zero. this makes sense -- if at any point the original signal has a value of Y, the polarity-reversed signal has a value of -Y. Add them and you get zero.
A very similar effect can occur from time delays (differences in phase). If you delay a sine wave by 1/2 its period and add it to the original, you will also zero out the signal. The wave oscillates from positive to negative. Shift one wave so that the negatives line up with the positives of the original, and it looks just like its polarity was reversed.
Since no two microphones can be at *exactly* the same place in a room, they will receive a sound at slightly different times. Sometimes this delay can cause different frequencies to be out of phase and cancel each other out. Low frequencies, with longer wavelengths, are particularly prone to phase issues. If you find yourself in this situation, and you reverse the polarity of one of the microphones, the two signals will now no longer cancel out. Problem solved.