I just conducted a little experiment.
I turned OFF ALL other processing except for MC, highlighted a short range of a song in Samplitude and Bounced down 4 passes of it as follows:
1- LINK set to 0%
2- LINK set to 50%
3- LINK set to 100%
4- LINK set back to 0%
I then placed all 4 files on seperate tracks of a new project, lined them up sample accurately and reversed phase & muted in groups of 2 and began playback (perform null check).
What I found was:
1&2- Resultant Peak level down to -136dBFs
1&3- Resultant Peak level down to -62dBFs
1&4- Resultant Peak level down to -156dBFs
2&3- Resultant Peak level down to -62dBFs
2&4- Resultant Peak level down to -136dBFs
3&4- Resultant Peak level down to -62dBFs
I definitely see a pattern but I have to say I don't fully understand it. I expected 1&4 (both passes with LINK set to 0%) to null completely; they didn't. Below the noise floor yes, but not a perfect null by any stretch. In every case, the passes with LINK set to 100% yielded the worst results with respect to nulling.
MC settings: No buttons engaged; Soft, RMS; FAT Off; Limit Off; Ratio 2; Attack 170ms; Release 284ms; Threshold -13.52; Tilt C; Mix 100%; Output 0.0dBFs.
PSP: can you shed some light...?