All this talk of the polarity of caps is kind of pointless. We're discussing audio signals here. By definition, they are AC. That means that no matter which way you orient the cap, signal is flowing "backwards" through it half the time. Please correct me if I'm wrong, with proof. I do like learning new things.
Signal is flowing backwards half the time, but the question is how does a polarized capacitor behave if signal is flowing backwards? Some of us are not hearing any difference at all, while I convinced myself that I did hear slight distortion with polarized caps. I also read that a cap becomes a short for voltage flowing backwards.
That's why I did investigate.
Turns out: A polarized cap is a capacitor regardless of the direction of current, but it basically breaks if the current flows the wrong way for too long. I think you can use polarized tantalum caps rated 12V or up for CIEMs without problems, according to the NASA paper that should be ok. You probably cannot use electrolytics because they will fail over time.
But of course you can always use two polarized caps in series with reversed polarity or unpolarized caps to avoid any of these risks.