The bias point drifts found in the common emitter configuration when no emitter resistor is connected, will not be relevant in a switching application like the one we discuss. In fact we can not talk about a bias point, since the transistor is not biased in the active region, but it works in two states (cut-off and saturation).

On the contrary, the bias point stability becomes very important in an linear application, for example an amplifier.

The use of a darlington pair (either discrete or integrated) increases the equivalent base-to-emitter drop, so degrading a little more the emitter follower option.

Regards,

Alfredo del Rio.