People notoriously won't know what a new badge means, and this being an uncommon scenario, no one will, panic ensues.

It might be better if such email was automatically marked as spam until the user marks it as good. If they keep doing that, then that send path should gain elevated trust. (I mean, they are using some ML system here right?) I'm imagining an anti-spam-campaign supervisor system that also tracks spam with some sort of soft hash (or ML), running across many user's email.