1. use to improve linearity. it is an easy way, low cost, and possible to implement degeneration without high common mode reduction if you use 2 tail current sources for Q1 and Q2, and the degeneration resistor is between them:
2. as you said to improve CMRR and keep consumption independent from Vcm
3. it is called quasi-differential pair, low supply voltage usually a reason why people implement it.
So use scenario 1 when common-mode voltage varies in any way (DC or AC), use 2 when Vcm and CMRR is not relevant, scenario 1 if Vcm is constant and VDD is very low.