I suspect these statements are wrong.
As we know, duplicate ACKs play a significant role in TCP congestion control. According to what you said, TCP would respond with RST upon receiving a duplicate ACK. That is simply not the case.
There are more reasonable use cases to contradict the following statements.
If the ACK isn't within the window, the host will respond with a reset to indicate that it has no notion of there being a connection with these "keys", where the keys are source, destination, source port, destination port, sequence numbers.