I was thinking along the lines of the second scenario where the labels are both the target and vector of abuse. Moderating what labels are allowed opens you up to claims of censorship again.

People love to misuse tools meant for good, on Reddit I've been on the receiving end of the "reddit cares" self-harm notification because of some barely spicy comments.