CDATA sections are pure syntax sugar, nothing more. This is discussed in Item 15 of of Effective XML,
"Build on top of structures, not syntax" <ejb-ql>
<![CDATA[WHERE (name = ?1) ORDERBY shortDescription]]>
</ejb-ql>
is exactly equivalent to
<ejb-ql>
WHERE (name = ?1) ORDERBY shortDescription
</ejb-ql>
In this case the CDATA section was completely pointless. However, in the general case it might be used to let authors include < and & characters without escaping them. I suspect some tool here is simply wrapping all SQL queries in CDATA sections rather than take the time to figure out which if any characters need to be escaped. That's all.