I also don't see the BJT helping. It's a follower topology so the output becomes [opamp out] - Vbe. It doesn't reduce the loop gain in any helpful way but insteads add more delay/poles.
I don't see how adding in a BJT is easier than a couple passives.
What do you have? What's the meaning of post #18 in this regard, the BJT buffer has been newly introduced to this thread, apparently it's not on your PCB.
The existing PCB can be e.g. used with mentioned LM7321 as drop-in replacement. It's also expectable that LMC6221 and many other general purpose OPs are stable under certain capacitive load conditions, unfortunately the behavior can't be predicted based on datasheet information.
What do you have? What's the meaning of post #18 in this regard, the BJT buffer has been newly introduced to this thread, apparently it's not on your PCB.
Thanks, what we actually have on the pcb is the schem in #18 [but the 2k7 (R5) is actually a 100n capacitor]
...we are currently wangling it to be as the schem in #17 (by not fitting the bjt and using a zero ohm between some of its pads, and not fitting the two 100n caps on the output)