Currently, timeouts in NAKACK are static, e.g. 300,600,1200.

The goal is to make the retransmission timeouts a function of the actual average retransmission times. Ideally, we maintain rolling averages of actual retransmission times per sender and adjust the retransmit timeout based on these values.

This would dynamically slow down retransmission on heavy traffic/congetion/message loss etc, and speed it up when those problems disappear again