Ok, I’ll try to answer to your questions.
This language is a proprietary program used to perform corpus (textual) analysis. It’s an expansive piece of software targeted to just a bunch of scholarly individuals. Even though everybody is used to XML in this field they decided to use a “simplified” syntax and even went as far as developing their own “linter”.
As far as I know their is no clear syntax definition either backus-naured or informal. It’s just “hey, the anchors/tags look like <TYPE=VALUE>”. Actually the manual of the program uses Comic Sans MS as title font.
Now, this syntax is indentation insensitive. There are examples in the manual that feature such a behavior. The scope of type-X tag ends where another type-X tag begins.
Concerning colors, the tags (both type and value) are displayed in red (or whatever color I might use), the actual text is fine (white). I was wondering if their was some convention in the way to name patterns so that they get displayed in different colors, but after some careful examination I think I might be able to handle this on my own.
Back to the folding heuristics : I actually got it working as I wished on one, clean, well-indented file but not on some indentation-less file. After some tests it turns out it works ONLY IF it is indented, so I suppose i must be relying on some sublime-text/textmate specific heuristic since I did not specify anything about indentation in my tmLanguage file.