If you can figure out what the old encoding scheme was,
you should be able to dump the defective items, convert them to the proper encoding, and reload them.
A thought occurs to me - you can use an ETL utility (like Pentaho DI/Kettle) to setup this process without having to do any arduous coding (after all, that's what you want to do: ETL - Extract, Transform and (re)Load). Dump the offending items out as text files (make sure you have a multi-lingual text editor!) using different encoding schemes for the source until you get something readable, then you can create the part of the pipeline that reloads with the repaired values. Needless to say, you should backup the database first - or better yet, clone the database and do your experiments on the clone copy.
The secret of how to be miserable is to constantly expect things are going to happen the way that they are "supposed" to happen.
You can have faith, which carries the understanding that you may be disappointed. Then there's being a willfully-blind idiot, which virtually guarantees it.
