Most detection products fail in the same way: they announce an incident and hide the reasoning. The operator either trusts it blindly or ignores it. Neither is acceptable in a control room.

TfGM SNV records the observations, the threshold, the window and the confidence with every detection. The console shows them behind a single "Why this fired" control. If the operator disagrees, they can dismiss the detection with a reason, and the reason is part of the audit trail.

That is a design choice, not a model choice. It is also what makes a rule reviewable after an incident: you can see what the system saw, and change the threshold with evidence.