It's likely that drain current is just the "most
accessible proxy" for gain, dead-flat-in-all-
dimensions is not a likely outcome but you
can at least find a place that delivers the
best minimum across-it-all forward gain.
If you understood all the dependencies and
had the sensors, you might make a more
complex bias control (but "good enough"
has many other virtues like cost and time
to complete, while "ideal" is waiting for a
bus that never comes). In the middle,
engineering.