Gain drift of a standalone op amp comes from device
aging / degradation (if you're talking drift over time and not
some environmental-stability spec).
A programmable gain amp, if it's discrete steps, adds some
other sensitivities such as resistor and switch resistance /
leakage drifts.
You could go backwards from lifetest data to see what else
has tracked, but I would not put much faith in predictions,
fellow-traveler-wise, until I trusted my ability to model
the individual elements' behavior over time.
Many op amp designs depend for their gain, on at least one
node being a very high impedance. That means no leakages
that have a resistive nature to them, can appear or increase.
Open loop gain is one of the most sensitive params, to small
device leakages (as might accrue from oxide charging,
mobile ions and so on).