Marc Daniel Nelson treats the entire vocal as one element through a stereo bus. Instead of a traditional point-filter de-esser, he uses Soothe 2 in broadband top-band mode to find and tame the most aggressive sibilance. Soothe scans across the high end dynamically rather than locking onto a single frequency. On vocals with varied sibilant energy, that sounds more natural.
Once the de-essing sounds right, he backs it off by 10 percent. That is his rule: when you think you have it, pull back a little. You can always add more later, but you cannot undo over-processing once it is backed into the mix.
He also flags that the top end of the vocal, the snare, and the hi-hats all live in roughly the same frequency space. Any top-end lift on the master bus opens all of them up together, so de-essing decisions made in isolation have consequences further down the chain. It’s always worth thinking about how elements share frequencies before processing.
He closes with putting a light limiter on the mix bus, set just to catch transients and prepare gain structure before mastering.