There was a nasty bug in ad_html_to_text which lead to funny characters and possibly runtime errors. ad_html_to_text was broken, when it encountered (valid) entities like e.g. & the problem was that the old code interpreted the numeric value of the entity as an octal value because it starts with leading zeros. HTML 4.01 allows only decimal and hexadecimal numeric character references.
This bug is fixed in CVS head and in the oacs-5-3 branch.