It sounds like they're doing multiple takes (which will all sound different) and trying to mix them all in.
Here's my approach to get a better mix than that.
1. Record two to three takes of the verse
2. Keep the one that you like the best
3. Run a de-clicker and a de-esser on that track
4. Make 2 exact duplicates of that track
5. Set them to start about .045 of a second later than the original track
6. EQ most of the bass out of one of the tracks you copied and EQ most of the treble out of the other
7. Pan one of them left and one of them right. You don't need to pan hard left and right, but you can experiment with them and try it out.
8. Once you do that, then you can do the same thing with your ad lib tracks and hooks.
Note: I also used Antares Mic Modeler and modeled my mic patterns to different mics. That's up to you if you do that.