Great question. There are a few ways we could handle it, but in principle we want to see how the problem happens before solving it. A confirmation process is a great idea but adds friction, and as we roll out we don't want to discourage usage. Basically, we want to see what happens so we don't optimize or problem-solve too early.