I have no say in the matter. 
But if I were designing the feature, I’d want to make it a toggle that mods can turn on and off after a message is sent. I have sent a message intending to be an official warning and forgotten to check the box. You’d probably want limit it somehow (only one change in status, perhaps) to avoid moderators abusing it.
I’d also want to consider how the change is communicated to a user. I think ideally the user wouldn’t know this happened if the mod fixed their error immediately. Not insurmountable, but the sort of complication that makes seemingly easy changes harder than they appear.