I just realised I forgot this in the help rewrite this week!
When you have a download page open for danbooru or whatever (if you haven't seen this before, go F9->download->gallery->whatever), there should be an import options - tags collapsible panel. Hit the 'expand' button and you'll have a list of your tag services and the namespaces that hydrus knows how to parse. Set these checkboxes how you want and as the page downloads each file, it will parse those namespaces and pass them to the respective tag services.
If you have existing, imported files that you want to tag with a site's tags, that's more complicated. There's an explanation here:
http://hydrusnetwork.github.io/hydrus/help/advanced.html
Under 'tag archives'.
I have decided to work on tag processing a little more this week. I think I'll split it up and push it to idle time, when the user isn't using the program, so it doesn't slow their gui experience so much. I'll be able to make it run a lot faster if I don't have to worry about trying to give the gui time to respond to user actions, as well.
Furthermore, I think with it set to sync in one long go, the db file fragments so much that it slows down hdd access. I suspect that hydrus either vacuumed or your OS defragged between your updates, which is why it is faster. I didn't touch the processing code last week, so it can't be that. Maybe there is some memory bottleneck stuff as well. I'll keep looking. Thank you for your feedback.