We tend to sometimes hold multiple simultaneous (or at least near-concurrent) conversations with different headmates, so latching that's global across the entire platform would be useless to us. Frankly even pluralkit's per-server latching is a bit annoying in some spaces. I'm not sure how hard it'd be to implement, but if we could choose the scope of latching, that'd be really useful.