Something appears to have changed as of 1.9 (although I'm not 100% sure exactly when the change happened) however as we're no longer receiving the hits we expect when using this work-around. We still get results as expected when using queries containing only "normal" ASCII characters, but the "slugified" queries no longer appear to match anything.
We've tried removing the work-around as well, and still get no matches when using Swedish characters.
One thing I've noticed that may or may not be useful is in the saved query files under _cache/_add-ons/bloodhound/searches. The query parameter gets HTML-encoded ("spärr" becomes "spärr", for instance) but in the actual file contents the text is still utf-8 encoded ("spärr" is still "spärr") like on disk...
Might something have changed in Statamic that Bloodhound simply hasn't been updated for yet?