Can you help me clarify some of my doubts as stated in the post?
The idea is that if you already have a random entropy you can derive another random entropy from that original entropy by performing certain cryptography functions on it (basically hashing it). For example SHA256 of 32 byte entropy produces a random but deterministic 32 bytes.
Although the algorithm used in RFC6979 is more complicated than that but the principle is the same. You have you private key (which should be random) and the hash of the message that you want to sign (which is
not random) and by performing a series of HMACSHA256 you derive a new key.
Since part of the input to HMACSHA was random (the key) the result can not be guessed ergo is random.
The benefit of this is that the algorithm avoids using any kind of RNG that can be weak and since it is deterministic, the ECDSA itself can be easily tested which makes implementing it so much easier.