My approach would be to add a DC stabilising path around the whole opamp, then inject a signal and look at the output from each stage, compared to it's main or stage input.
eg. as a concept, a 100K from output to a cap to mid supply, then 100K from the cap to negative input.
Bias the positive input to mid supply with another 100K and capacitor couple the test signal to one of the inputs.
Add a cap to ground from the other input if needed & lower value resistors if the input currents cause too high an offset.