While I agree with the results, I don't agree with the explanation of question 52.
This scenario doesn't happen:
Imagine the following scenario: t1 starts, it invokes aMethod. i is now 1. t1 releases the lock associated with a and the scheduler decides that it is the turn of t2 to run. t2 invokes aMethod and i's value is now 2.
The code actually creates two Question52 objects, each with its own A object.
The instance methods aMethod() and bMethod() synchronize on the current instance of A. Since ++ and -- are not thread-safe, threads t1 and t2 could easily corrupt static variable i during their combined 20,000,000 executions.