Yep. Me too, as I can't even read that faecsebook page...
I feel with you. My FaceBook page has been compromised last week and all posts and friends--including Steve--are deleted. -Sigh-
As for how it was done:
A first glance at the project rate (192 kHz) suggests that something is hidden in the upper frequencies--Steve is usually not a fan of high sample rates.
Essentially, the hidden tracks could have been added directly in the frequency domain by stacking up 4 FFTs on top of each other.
Looking at the spectrogram reveals two prominent peaks at 30 and 60 kHz.
This excludes spectrum stacking because the energy would be differently distributed.
Therefore, the audio must have been modulated.
I tried FM demodulation, et voila, it worked.
It is good that it wasn't FSK, PSK or something of that sort...
Or a 24 bit audio with three times 8 bit tracks (two hidden as inverse quantization noise).