If they did, I'd have to maintain a bunch of different plugins, one for each server I want to support, across all of their different languages and plugin frameworks and APIs, on top with the bridge code itself. Not nearly as feasible as the current plan of integrating at the protocol level, and making the bridge just another instance, and bridged users normal fediverse users, so that users and admins can limit/block both users and the bridge as a whole like normal.
And yes! Bluesky moderation labels are great, I think that framework probably has more long term potential than the fediverse's current instance limit/block approach. @thisismissem is working along similar lines with her FIRES and Fedicheck (for IFTAS) projects.
Moderation labels don't really translate well to hashtags specifically, but the bridge will absolutely support Bluesky labels in general. The Bluesky team's AppView already filters out most of them (including all objectionable ones) by default, and that happens upstream of the bridge.