This is more tricky than it seems when coding it absolutely right. Overflows can happen due to spikes caused by the trigger. Those should ideally not be taken into account. But also allow overflows in X% of traces (say X < 10) vs allow overflows in X% of a single trace (say X < 0.1)?