Yep Mark, it's not at all uncommon for some to use MBC as a de-esser. You are correct in that you'd first identify the frequency that is the offender. You'd then set 1 of the bands to center on the problematic frequency. You'd need to narrow the band. The threshold would then need to be high. The attack and release would have to be the quickest available and adjust from there if needed. There are some downsides, however, particularly compared to a decent dedicated de-esser. Both of the MBC's I use (Izotope and McDsp) do impart a sonic signature. The McDsp much more so than the Izotope. I like the signatures but that would be inappropriate for a bunch of the dialog I work on daily. Also using a MBC on a vocal/dialog track that had several problematic frequencies would obviously require multiple bands. At that point, it might well be too invasive to the sonic as a whole.
The way most, in my world, de-ess is to use a destructive de-ess plug-in. In my case, Pro Tool's destructive Audio Suite protocol works great. The plug-in first allows one to reverse the sound so all one hears are high frequencies. You sweep the frequency until you clearly hear the offender. Set the de-esser to that frequency. Then reverse the plug-in back to full frequency. From there I zoom in on the problematic "s", highlight them, and destructively apply the de-esser. At that point, the only thing that is affected is the singular problem "s". Also, since it's easy to reset the frequency point I can address any problems at any frequency that is offensive. It's certainly more time-consuming than just running a MBC or de-esser in real-time, but it is surgical, non-invasive, and leaves the vocal/dialog sonically untouched.