Agreed on Q4 — if understanding race conditions is what Colin was driving at, Q4 was simply a bad question for it.
A better wording might be: "What surprising behavior could occur if `pthread_cond_wait` didn't take a locked mutex?"
(Full disclosure: I also scored a zero on Q4. However, I'm familiar with both pthreads and the necessary atomicity between releasing the lock and waiting on the condition, so this was a surprise. He was obviously fishing for a specific answer, but I couldn't figure out which from the question.)