Applying compression during mixing is the wrong solution, it damages the source material. It would be better to provide compression controls in the playback device.
However, in modern times, most wouldn't want to manually adjust such settings. A possible solution could therefore be something akin to the compression controls or dynamic eq already seen in AVRs.
Thinking out loud here... Further down the line I wonder if we might see something similar to Dolby vision. I.e. HDR Audio, where metadata accompanies the audio stream with flags to specify the amplitude scale that the DAC should use. (This could allow more optimal use of the available number of digital bits).
...Compression control could easily be implemented simply by adjusting the transfer function between the metadata amplitude flags and the amplitude scaling. (Equivalent to dynamic tone mapping of the Electro-Optic Transfer Function EOTF in HDR displays).
Edit: On second thoughts... Perhaps HDR audio doesn't make as much sense as HDR video. Most displays have only 10 or 12 bits per colour channel, but have millions of pixels per frame. (The bit depth is a limiting factor for the visual appearance, but cannot easily be increased due to the enormous bandwidth needed). For audio it is already practical to have 32 or 64 bits per channel (typically only 10s of channels required, maximum). ...At which point there is no need to mess about with flags to switch between amplitude ranges. (16 bits is probably already enough for most audio distribution and 24 should easily be plenty).
