I'd say that the reason why the parser does not complain about line 8 is that it has no valid reason to say an ERROR element cannot contain another ERROR element.

In order for the parser to know this you would have to use a DTD to describe the valid document tree.

Consider this syntactically valid set of tags:

<ERROR>
  <FOO>
  </FOO>
  <ERROR>
    <BAR>
    </BAR>
  </ERROR>
</ERROR>
[download]

Now, if you remove the last line, the XML becomes invalid. It would be tempting to fix the problem in line 4 but this would give a completely different document structure.