Taking two timestamps (one before the function-under-test, and one after), we minimize the chance of failure.