This is a bit of a tricky question, to be honest. Digital audio is processed based on it's sample rate (frequency range) and its bit depth (degrees of difference between zero amplitude and full amplitude). Currently, digital audio can have three bit depth options: 16, 24, and 32-floating point. When analog audio is converted into digital samples, it quantizes the fluid motions into rigid steps of volume. For example, 4-bit audio has 16 different layers of volume:
16 bit and 24 bit audio have an upper limit on the decibel level of information they can contain, which is considered 0 dB. Any information processed above that limit is cut off, and is not reproduced by speakers or on a bounce of the processed audio. So when you are working in 16 or 24 bit audio resolution, it is incredibly important to properly gain stage everything and avoid redlining without specifically looking for that distorted sound. So the rules on headroom came about as a preventative measure against unwanted digital distortion.
The difference in steps of loudness between bit depths is not linear, it's exponential. There are 65,536 possible quantization points in 16-bit audio, and there are
16,777,216 for 24-bit audio.
32 bit floating point audio treats things differently. At its maximum, 32 bit audio has over
4.2 billion steps of quantization. But not everything needs that much. So it starts at 16.7 Million, and it adds on more points as needed. This allows it to process not just information that is recorded at redline, but information that goes
over zero decibels. If you rendered out a 32-bit wav file that went over 0 dB and normalized it, the file would actually get quieter as it brought the highest peak to 0. You wouldn't lose that information.
What this means is that if you are working in a DAW that utilized 32-bit floating point audio processing, you can have audio that clips without it necessarily being a problem. You're not LOSING the information... inside your DAW. Your interface is 24-bit (there are 32-bit interfaces out there but they are still fairly rare). Your speakers are 24-bit. Compressed audio files are 16 or 24 bit. So as soon as that clipped audio goes out of your master channel, it is subjected to the same sort of problems as if you were working in a lower bit depth. As long as you are in some way preventing the master output from going over 0 dB, you have nothing to worry about.
Personally, I take this as a sign to just get in the habit of accounting for headroom and trying to make sure my track plays well, while giving me much more wiggle room while I'm working towards that point.
