"I've notice this, and wondered why. It may, indeed, be a math issue in the code, where larger numbers have a disproportionate likelihood of suffering damage or loss. For example, if you move an infantry division from Oran to Naples in short bursts, you have a good chance of getting all the way there without naval attrition, whereas if you do the entire voyage in one click you almost certainly will suffer attrition. I believe one of the designers confirmed this to me. The same might be true for 75-aircraft air units v. 16-aircraft units." Cary
What was interesting to me was that the 75 plane units were suffering more morale loss even with proportionally lower combat aircraft losses and damaged planes. The morale hit to the P-47 groups was thus harder even with miniscule losses. I was losing 3-4 groups of P-47s in 9th AF per turn to morale. Make them 'squadrons', take the same combat losses and lose almost no squadrons to morale & when you do have to rest one squadron it is fewer planes off line. I will try the shorter burst naval movement - maybe the same issue, a simplification in the math of the probabilities in the code.
The game coding problem is probably something like this: the designers want the game to run fast. But, floating point math and especially division is slower than integer math. So these issues are at risk of getting into games due to the possible use of integer math where floating point would be more accurate.
The Python script works on the air group csv, then reassigns the pilots in the pilot csv. The editor seems to sometimes put too many planes in a 'squadron' when it is closed, more than the number of pilots, but the game also seems to correct it at turn 1 end. You wind up with unused planes overloading some bases only on T1; but as they don't seem to get used they don't appear to do any significant harm.