When using items from more_subfields_xml in item search there are performance issues, especially on bigger data sets. more_subfields_xml is an XML blob type column in the items table, that will be used when you create additional item fields not mapped to another column in the items table. Not sure if something could be done here, but maybe there is something we could do to optimize processing?
It looks like there is a general issue with item search being very slow in some cases. One issue is left truncation of search terms. The example is an ISBN search. Multiple ISBN are stored in biblioitems.isbn | separated with additional ISBD punctuation. So you'll need to truncate left and right to make sure you get correct results. Left truncation creates an issue as we assume the index on the field cannot be used in this case.