I have not had to deal with many hive precision issues, also I do not really have an opinion on how things should work. But I think we might want to consider deferring to how Java does things rather then how MySQL handles things.
On one had its nice to handle things like the mysql reference says to handle things. Of course the mysql reference will have from subtle to not so subtle changes between versions.
While I think it is great to have all the UDF's that mysql may have, I don't think we should always defer to doing something as mysql would, unless mysql is doing it to match some larger SQL standard.