Yes, it seems a bug in the Qt4 API, it should be documented that this is 'too fragile" to mess with.

I am toying with the idea of using the syntax highlighter code to know the state of each word in a paragraph. This will help me (for example) knowing which words can I send to a spell checker (only words inside comments or strings for example).

Yes, the variable I showed before only gives me the state in the beginning (or end) or each paragraph.