Its a *LONG* lime since I studied that and it was only mentioned in passing, but IIRC it needs to have a large and reasonably linear change of Hfe with Ic so its gain can be controlled by varying the bias point. The MMBT2904's gain curve is essentially flat between 0.1mA and 20mA so it would be a poor choice for this application. Also, you'd probably need to specify a manufacturer as the shape of the gain vs Ic curve may be rather different for a second source part NOT designed for forward AGC operation.
This doesn't sound quite right. Hfe is the current gain, and it is typically fairly constant over a wide range of collector current.
On the other hand, the transconductance gm is roughly proportional to collector current. This is true of every bipolar.
Gain is mainly determined by gm, not Hfe.
So just select a part with similar beta and Ft.