# TCP delayed duplicate problem

The TCP(Transmission Control Protocol) delayed duplicate problem is discussed in Tannenbaum's "Computer Networks." The discussion references Tomlinson (1975) improved by Sunshine and Dayal (1978). Basically, from what I could make out it uses a ToD (Time of the day) clock to uniquely number packets and ensure that, if packet lifetime is $T$ secs, then no new packets have the same sequence number within $T$ secs. To elucidate the above fact a graph between sequence numbers and time is drawn which shows the linear relation and marks a Forbidden region, entering which produces duplicate packets.

So here are my questions:

1. How exactly does entering the forbidden region produces the delayed duplicates?
2. If I wanted to quantify the discussion with the following parameters:

a) ToD clock size - 32 bits(keeps running even when the host is down)

b) clock increases once per milliseconds

c) maximum packet lifetime - 64 s

How would I find the minimum and maximum permissible sequence number generation rate?