On using SyntaxNet's Parser with LT

Recently Google released their trained English Language Parser, Parsey McParseface .

https://github.com/tensorflow/models/tree/master/syntaxnet

I am not sure how correct will the Parser be to get the linguistic structure of grammaticaly incorrect sentences.
Do you guys think, we might be able to define more complicated grammar error detection rules using LT with SyntaxNet’s parser output?

If its part-of-speech tagging is accurate, I’d love to see it in LT, though it might be problematic since it is a python project (vs. LT in Java). Anyway, if someone could make some comparisons about the performance of the current LT tagger and syntaxnet, it would be most informative.