For high DC precision one wants low temperature gradients and thus low power. So 4.5 mA, especially at higher voltage are a real downside. One would not like 9 mA at some 30 V and thus some 270 mW from a part for the 100nV range. Using 2 chips would be a small price for half the local heating.
The ADA4522 is at some 850 µA only and thus about 1/5 the power. Even there the heat is a downside for using the dual version. There could be a small advantage (but also some disadvantage) from having only one frequency in a dual chip. The high current OPs listed are more like low noise, but not precision types.
For the current noise, I somehow got wrong data ( ? old data sheed) in my table for the ADA4522
. It also has quite some current noise and the cross over is at some 45 K, where the ADA4522 would give similar total noise.