We used an auto siphon and a bottle filler to transfer the middle layer of the "less oil" jar into a pint glass. I held the end of the auto siphon in the middle of the layer, and then slowly operated it to push liquid into the glass. There was some disturbance of the bottom layer noticeable during this, but not much. Some oil did end up in the pint glass, but it did not appear to have any extra sediment (same color / cloudiness as in the jar):
Most of the oil appeared to stay in the jar:
Based on these results, my current plan is to proceed and use the full peanut in the beer. I believe that siphoning the wort from the boil pot to the first stage fermenter will remove a lot of oil, and that the second siphoning will remove the vast majority of the rest. I'm confident that if I need to, a third siphoning will clear up any little that is remaining.
I've placed the jars into the fridge to see how cooling them affects the separation - if it is winter this is a viable option for the carboy / fermenter in the full scale batch.
I'll probably try one more small scale experiment where I double peanut butter to water ratio, to see if I can increase the peanut butter flavor but still manage the oil.