I'm not sure which comments go with which schematic, but several things.
On the diff amp, C75 should be across R161.
The very large resistors in combination with the op amp input capacitance gives another pole/phase shift.
For the non-inverting, R267 is unecessary with picoamps of input bias current.
The 10 ohm Riso is probably too small to isolate 1 nF.
Thanks for your feedbacks, I couldn't agree more on your improvements you proposed.
Even if my oscillation issues had been stopped for a long time I had never been able to replicate the oscillations on Spice simulations. Even worst, a stability analysis shown up that phase and gain margin would be acceptable.
Would it be possible that high impedance resistors have a higher influence on stability issues on amplifier inputs which has auto-zeroing and chopper-stabilized features like AD8629 than other usual amplifier like AD822 ?