lucy24 - 12:47 am on Sep 22, 2012 (gmt 0)

If it's database contamination, why are the older entries readable?

Because you changed horses in midstream. You can either tell the database to convert from ISO-8859 to UTF-8, or tell it to keep UTF-8 throughout. But unfortunately you can't tell it to treat one part of the database one way and another part the other way. There is no alternative but to go into your database and fix it manually. Is there a mechanism for splitting one database into two, and merging two databases into one? Any way to dump them into a text editor, do some stuff and dump back?

I know what you're going through, anyway. I've had the equivalent thing happen in e-books when I didn't notice in time that I'd opened the raw file in Mac-Roman and then started adding non-ASCII characters in what I thought was ISO-Latin-1. The only comfort is that it was a text editor, so "all" I had to do was globally replace all the garbage pairs, like your å, with the proper letter.