Please explain "parallel smear/coloration". You talking EQ?
transformer modeling, old neve console modeling, tube saturation.
All of these
can induce a variety of coloration, harmonics, and soundstage smear (where you have more trouble getting a
pinpoint vision on stage of where an instrument sits in space). For example, the sonosax induces smear (among other things) on the soundstage, it also widens it, but it smears it so you have more trouble pinpointing in a razor sharp way where an instrument is. If a trumpeter is at 10:30 in your mix, with the sax or other gear that induces soundstage smear, you might say they were somewhere between 10 and 11. You know it's forward and left, but you couldn't close your eyes and pinpoint
exactly where they are. That in trade can create euphoria so sometimes it's wanted. One instance where smear didn't bother me as much was in PA taping (what is your soundstage; it's 2 stacks and a bunch of drunks. I don't necessarily need to know it's 2 stacks that are 30 degrees in front of me). One place where it irked me was taping stuff on stage. If you like that razor clear soundstage, smear is evil. If you want euphoria in your soundstage, then it's a valid trade.
So take that concept and now think about running a parallel comp and then using something to smear or color that set of tracks and mix it back in with the non-colored/compressed. The theory** being that stuff would pop out when they surge or solo and would smear when they don't. Another example would be if the band was quiet and the drummer did a sharp cymbal crack. You would hear the crack off of the non-comp/color track and then hear your coloration on the decay.
** This is a theory because different plugins have an inherent delay in rendering than others do. For example, the Softube FET compressor delays it's output from the plug by 44 samples, so I can't just mix that back in with the original signal (because it sounds like ass, I tried), I have to delay the unprocessed signal by 44 samples and then mix. The URS Saturation plug doesn't seem to. I measured by creating a single sample spike in an otherwise silent 2ch wav, processing one channel and counting sample differences afterwards. The theory part is did you measure it accurately. The more accurate the measurement (even down to half samples or quarter samples if your DAW will handle that) the better the result.