I just want to provide a little background for anyone confused by all this clipping and attenuation nonsense. :-)
During the decoding process, PCM samples are reconstructed from MPEG audio frames. The nominal output range of the PCM samples can be considered to be a fraction number between -1.0 and +1.0. This is eventually scaled to a number with certain resolution, for example 0..255 for 8-bit output, or -32768..32767 for 16-bit output, etc.
Because of the nature of the decoding process, it sometimes happens that PCM samples are calculated to have a value less than -1.0 or greater than +1.0. This isn't really an error, it's just an artifact of the lossy compression and depends a lot on the encoder and the strength of the original signal.
Most decoders have little choice but to clip the output signal to the maximum representable values. This adds distortion to the signal.
By "attenuating" the signal, the entire signal can be weakened proportionally so that the peaks fall exactly on the -1.0 or +1.0 full scale limits rather than beyond them. It is a digital volume adjustment to prevent clipping.
Internally, MAD represents signals with about 169 dB of dynamic range and enough headroom to calculate samples 18 dB above full scale. It's therefore possible to further manipulate the signal without losing quality before it is finally scaled and dithered to the output resolution.
The "Attenuate" button in the pre-release activates an attenuation filter to weaken the output signal based on the amount of clipping experienced. An automatic version of this would do the same thing as soon as clipping is detected.
Cheers, -rob