In WordDetect rules, verify delimiter characters also on the right and left edges inside the string.
For example:
<WordDetect attribute="Keyword" String="<hello"/>
In the past, this rule was equivalent to \b<hello\b in regular expression. Now, it's equivalent to <hello\b, since < is a delimiter character.
I have checked the WordDetect rules in all definitions and and I haven't seen regressions. In the definitions elm.xml, selinux-cil.xml and selinux-fc.xml I replaced some WordDetect rules with StringDetect, since in this change they are equivalent.