@dynamiclinearity said "Power is used for very, very short term transients which can be 20 to 30 dB"
This misrepresents what is actually on recordings. Over two decades, I transferred over 1,000 LPs -- from rock to classical -- from LP to digital format, so spent a lot of time watching loudness levels during the transfer, using Adobe Audition. I've never seen a recording where the average loudness level allowed for 20 or 30 dB peaks.
Now, lots of recordings have "dynamic range" in excess of 30 dB, but that includes the soft parts -- perhaps a solo instrument being softly played during an orchestral recording. But, the quieter parts of a recording are supposed to be played at a soft level. Those soft parts are well below the average level.
Then there are rock recordings where one often sees the average level almost at the max level permitted by the media -- one may be lucky to have 3 dB peaks in such recordings -- remember the loudness wars?
Simply put, if your average recorded level is -6 dB, there is simply no place to put a 30 dB peak transient. If it existed in the live music, its been lopped off in the recording. LPs are slightly different from digital recordings, but they also have their limits, too.
In short, to have a 30 dB peak transient on a digital recording, the average playback level would have to be 30 dB down. There aren't very many recordings that meet that criteria.