With the new index search system, It seems nearly impossible to search Japanese words (2-byte character).
In version 2005-07-13, I could search Japanese words properly, and in 2005-09-22, I can't.
I think new indexer recognize only words which are space-separated (or separated by beginning and end of line) or used as title in links or headings. In Japanese language, spaces are not used as word-separator, so most words in documents are not searchable with new search system.
Paradoxically, search for pagenames in 2005-09-22 becomes better than that in 2005-07-13. In 2005-07-13 version, Japanese characters seemed to be allowed as pagenames and namespaces and it worked well, but Japanese characters in pagenames were never hit in search. In 2005-09-22 version, Japanese characters in pagenames are now hit precisely in contrast to those in documents becomes unsearchable.
I think the easy way to solve this problem is to create some option to choose old search system or new index search system.
Otherwise, Morphological Analyzer for Japanese language like MeCab, Chasen, Sumomo and Kakasi or N-gram will help.