[[Nocc:C780de0d3ac3c245b859f5a2b8e4782ec30f8e3d-Y-m-d/qqq]]
Unsurprisingly it breaks if you add html to places where it is not supported.
NOCC Developers should find a way to fix this. There are places where you cannot do without correct language tagging (which I believe might be the cause of trouble here) and when you reuse a message (which seem to be done here) you must take care to keep markup in place where it has to appear, and strip it, where it must be not (inside html attribute values for instance) - after all, either is easy ;-)
There is currently exactly one NOCC developer. The website is also in the subversion repo. If you really, really care, please have a look at the code and submit a patch. I'm certain he will apply it.
Is this language tagging really needed and which benefits has it for the user?
For example text to speech systems can use the correct language.