The reasoning for how we know this:
If A is normal, then we can get a yes or no for the first answer no matter what $S$ is. Because we don't know anything about A's identity after the first question, the answer we get for the $S$ we ask about must be something that either a knight or knave could have said under some circumstance. Thus, the answer to the second question must allow us to distinguish between one case where A is a knight, one case where B is a knave, and the two cases where A is normal.
For the (K)night, (N)ormal, and Knave (sometimes called a (J)oker), we get six possible arrangements and the following possible answers to the second question:
Arrangement | Q2 answer
K N J | K N J
K J N | K J
J N K | K N J
J K N | N
N J K | N J
N K J | J
From this we can see that if the answer to the second question is that C is a knave or normal, we won't be able to tell if A was a normal or not. So the answer to the second question must be that C is a knight. Either arrangement with A being a knight allows for that answer, so in order for "C is a knight" to distinguish between A being a knight or knave, we must have narrowed the possibilities down to one of the possibilities where A is a knight, the possibilities where A is normal, and the possibility where A is a knave and B is a knight. So "C is a knight" will always rule out A being normal and B being a knight, leaving A to be the knight.