Right now the source audio is decoded, assuming it is encoded, compressed again and then sent to the headset, where it is decoded again. Newer codecs work a lot better than old but you still have two iterations of lossy compression. That said queuing up many songs in some sort of buffer doesn't sound workable. For one thing, many people stream from Pandora, et al. Then there is pausing, audio coming from the phone, and things like that (I am sure you wouldn't want your turn-by-turn navigation directions delayed for even a few seconds).
Back to the double compression thing; I would imagine Apple's solution here eliminates that extra round of compression which is probably going to be in the next version of Bluetooth anyway. Technically it has already been there for some now as you can have a bluetooth device do the decoding of MP3 and other formats, as long as both devices support it. In that scenario only the raw MP3 date would need to be sent, eliminating the double compression. But hardly anyone does that, and of course it would not work with DRM.
Mike
Audio DRM seems like the least likely thing to implement with this move. While it would deter casual piracy, it would not be hard to defeat. The minute the decoded stream is converted to analogue it is pirate-able, and considering the potential for higher quality headphones with better DACs and amps, not to mention adapters, the end result is going to be better than than what is currently output from the iPhone's built-in DAC and compromised amp.
It's not like video which originates in the visual realm and stays there.
I suppose that the transducers themselves could have some kind of digital feedback chip which receive the DRM signal as an ultra-high frequency pulse that authorizes them to play the audio, so when severed would terminate the reproduction, but that still won't stop just putting a stereo microphone between the ear cups and getting a pretty good recording directly from the speakers.