Usability Test – Boomerang Technique going wrong

I am doing a usability test very soon and I have some questions on how to make use of the 'boomerang' technique.

As mentioned here, when users ask a question during the test about some task, for instance:

Facilitator: Gives task (...)

User: Asks a question (ex. So what do I have to press for that?)

And the ideal response from a Facilitator is: "What do you think (you should do)?"

Which is great, but what if the user responds to this question by saying something like: "Well, you told me what to do, so how am I supposed to know?"

1) How do you deal with such situations when the conversation is completely blocked?

2) Also, if you find out that your task was phrased badly (even after pilot testing, you didn't know), and the user asks a question about the task, are you good to answer it? Or will this invalidate your findings?