So to combat use cases like this, why not just add a repeat option? There would be no break if it cached the beginning again.
The first two minutes are an ad, and having a loud voice talking to you all of a sudden in your bedroom while you are asleep tends to wake you up.
Also just download the audio you want and loop it yourself. It would take roughly 2 minutes and use way less bandwidth.
With compression techniques being as they are today, I truly don't even worry about the bandwidth.