There are still plenty of products on the market that use old converters that have horrible saturation behaviour. (Rollover, where it goes from max positive to max negative, is one of the most horrid).
The last converter I worked on (PCM1865) was defined very clearly to "clip like an opamp", so you'd clip at 0x0FFF (24 bit) until the signal comes back within range.
Not surprising, the customer expects it to clip cleanly or doesn't even think about it.
However, there are a few arguments to avoid doing this
A) Go much above the input signal range and you'll start shorting the signal through the device ESD diodes to the supply rail. Not good for your source
prudent design could limit over drive capability with design of analog side input stage.
B) THD near clipping increases exponentially.
not sure what you are describing. Some analog paths saturate fairly suddenly, some over a small range. I would expect saturation in a A/D convertor to be more sudden, than gradual. Note: there is one digital console with a soft limiter built into the mic preamp that some customers apparently like.
Given the extended dynamic range of mastering ADC's, you have more dynamic range than you have THD.
not sure what this means?
Better to sacrifice 2 or 3 dB of headroom and get both good SNR and THD.
that is the classic trade off between noise floor and saturation. For years operators with a little(?) digital knowledge hit the convertors hard to avoid quantization noise. This is not an issue with modern convertors that have an analog sounding noise floor.
C) Why commit distortion to a recording? Wouldn't you be better recording clean and adding distortion as a post process?
Mastering is the last step of the process where we stick a fork in it... In my personal opinion clipping is undesirable, but they aren't mixing for me as their paying customers.
JR