Re: "a brave person who said that HRA has no audible benefits"
You may misunderstand my context, 142... for a final output, surely, use all the bits you can (but be aware of the nastiness of compression-for-loudness).
When I say 18dB headroom, I would expect the nominal zero dB signal to be 18dB below 0dBFS - but I would expect that 18dB to be used by the signal.
In, say, a live studio environment you will find that live voice and music both have level extremes which are generally unrehearsed and, trust me, the last thing you want is digital clipping. In analogue systems the approach was to assume 12dB headroom since an analogue mixer is usually a bit kinder as the signal approaches clipping. However, even there, you are shoving a 775mV signal through amplifiers with at least +/-24 supplies.
In a music recording studio (as opposed to, say, a live concert) you would, I assume, have much greater control over individual levels. Nonetheless, you would generally want a sound mixer with at least 24 bit internal representation, even with a 16 bit input and output.
However, it's some years since I left the BBC so while I have a lot of experience with good studio technique, practice may well have changed since I left. And of course, music recording is a different kettle of fish from live studio work.