But that still doesn't explain how those samples which are restrained by the Limiter, and then CD medium can be reimported and read to have "clips"* [which I understand are signals going over a certain, defined level. Which the user would presumably set up, - so it could be -0.2dBFS which means that it's not actually clipping; just breaching a predefined level.]
In the digital domain, clipping is about the shape of the waveforms, not about the level of the samples.
Clipping represents an
attempt to exceed a certain level which failed. The waveform rises as intended up to a point, then fails to rise any further - result is clipping of the peaks. This is bad because it changes the shape of the wave - which of course defines its sound.
This change in sound will persist regardless of the final level. So a limiter which produces audible clipping artifacts limiting a wave to 0dBfs will also produce audible clipping artifacts when limiting to -0.3dB or -6dB or -23dB etc...
Using a plugin to detect equal-value consecutive samples is no use. Some limiters break up the flat DC peak anyway, so output no consecutive equal-value samples even when the peak is audibly distorted. And even if the peak
is flat from limiting but the maximiser then adds dithering and outputs the wave with -0.1dB ceiling the flat peak will not be flat once dither is added, and the 0dBfs maximum will never be touched.
Then there's inter-sample clips. That's down to how the hardware reconstructs the wave from the digital information, which is a different issue. Read the helpfile on SSL's free X-ism plugin for a clear explanation of how that happens.
X-ism when I tried it with a bunch of limiters consistently rang the cherries for any value higher than -0.3dB. Somebody on this forum (IIRC James Murphy?) posted a long while back that a guy from a pressing plant had recommended -0.3 dB as the maximum level for a CD, so I'd been using that anyway.