Ok, I tried a few tests myself here. A little different setup though. I wanted to make sure the unit would not get any supplies from the outside. Also, since in my game China is a lost cause, I'm personally more interested in seeing what I can do to a Japanese unit. So my set up was:
take Coral Sea scenario, modify it to add several bomber groups and an HQa to Noumea. A Japanese BF was placed on one of the tiny islands near by. No base (not even a dot base). The BF consisted of:
10 75mm T88 AA Guns
24 Av support
12/36 Engineers (run with both to see how much additional supply was sucked)
22 support
1 sound detector
2 observer squads
The unit starts with 300 supplies in all tests.
For the first test, a control.
12 Engineers
Turn - supplies (on hand, not supplies required)
1-300
2-295
3-290
4-285
so it looks like 5 supplies a day, while sitting idle.
Test 2
12 Engineers, 1 x 12 B25 group ordered to ground attack, 6kft. 1 PBY Sqdn on recon
1-300
2-131
3-48
4-27
5-6
7-0
In each phase, all 12 B25s flew. Note that this is not a linear decline in supplies. Rather, the best curve fit is a log fit, though it is not perfect (R^2=.9412). This makes some sense as the manual hints that under supplied units conserve supplies. Section 15.0 says
While supplies are
actually consumed as used, without adequate supplies on hand to meet the expected needs,
units instinctively begin to curtail operations in order to stretch out the available supplies.
I think we are seeing this in action.
Anyhow, moving on.
an abbreviated re-run of test 2
and a short re-run:
1-300
2-206
3-111
so less effect. Similar order of magnitude, but it appears there are die rolls involved here.
test 3
7 x 12 B24 groups ordered to attack
turn-supplies-(A/C in attack 1, A/C in attack 2...)
1-300
2-26-(33,27,6)
3-0-(35,18,9,9)
so supplies drop faster when attacked by more A/C and / or more waves. The drop is too quick to get any kind of trend line.
The remaining tests add 24 additional Eng squads to the base unit for a total of 36.
For the first test, I tried to do a baseline... but I left the PBYs on recon by accident. This is interesting:
1-300
2-221
3-153
4-148
This surprised me... why would 24 extra eng draw this many more supplies? Oh, right... I left the PBYs on recon. Doh! It seems that the base force is firing at the PBYs (no surprise here... we know they do that). But it seems like they are shooting an awfully lot at a handful of A/C. My sense is, that if anything is broken (and I'm not yet convinced it is), that this may be it. It also MAY be the difference between what the moose and yak are seeing. Yakface, do you know if your opponent is sending recon over this area at all?
moving on again. This time the real baseline with 36 Eng
1-300
2-294
3-288
4-282
So, those extra 24 squads draw an extra supply a day. Because of possible rounding errors, I can't say for sure if that means the involuntary fort building is costing any supply. That said, it is either costing no supply or very little supply - so if it is a supply vampire, it is not much of one.
A few notes on the testing:
I only put the one base force at the location. All bombing was ground bombing from 6k ft. Since the unit was alone, it was the one targeted in the bombing, and suffered casualties... between 0 and 10 squads disabled according to the CR. I did not control for this in my quick test.
Some parting thoughts:
It appears that what Yakface sees can be duplicated (with the above caveats). However, I'm still not convinced this is a bug. I would expect a large supply increase in combat. I would further not want a unit to be able to store a full months worth of combat stores that moved organically with the unit (which would happen if the supply required number scaled 100% with supplies used) as it would make extended behind the lines operations possible without need for a logistical tail. It would also suck in all supply from an area and cause other units to starve while some got really fat.
If there is a problem though, it may be in the firing at recon (and possibly search) aircraft.
Another thought / observation is that a fat unit is going to act fat. Units appear to make a good attempt to conserve supply when the unit supplies are low - however, the unit does not appear to take into account the local supply situation outside the unit. It doesn't matter if it can't draw any more supplies, if its got it - it will flaunt it. I don't know if there is a way to selectively starve a unit - if there is, this could be a possible solution.