Ok, Lephturn, as I stated in my post above, I enabled Vsync, and ran some "Control" tests with the 20mm cannons on the N1k and the Mk 108. I got the exact same results.
As for online vs. offline, I also ran a control with the .50 cal online. In fact, I have destroyed several hangars in combat conditions with an m16 using around 1300 rounds of .50 cal. The big problem with online tests is that netlag and server lag make it very difficult to determine exactly when a building falls. Add to this a bug where firing in short bursts (so as to arrive at a relatively precise number of rounds fired) results in shots being lost, and frankly, the online environment has too many results to control.
Testing against aircraft is even more difficult, because you have two connections to deal with, as well as the host-side damage model. I concur completely that caution should be used when translating these figures into air-to-air lethality, because there could very well be other factors modelled, such as penetration, hit location and so on. Heck, it takes 25 AP shells to destroy a hangar, and 20 HE shells; but the former are much better against tanks. Then again, they may not be modelled; mk 108 30mms should be relatively useless against armor, but instead they're quite good.
In short, these results show:
A. The relative damage on hangars of all the weapons in AH, translated into equivalent pounds of bombs.
B. The ground attack efficacy of these weapons.
C. The existence of a serious and repeatable bug in hit detection, at the very least with hangars (actually at least two of them: a. in some places the hangars are 100% transparent, b. on the "solid" parts, some shots travel straight through -- and,no, I don't think it's just dispersion mixed with a.).
C. does limit the validity of the results, but since applying the "firehose" technique of shooting generated results that were 100% repeatable on my FE (in different A/C and vehicles, and regardless of whether I was at a large, medium or small field), I hope its effect is minimal. Of course, it would serious mitigate the results if, say, its effect were more pronounced with the number of projectiles in the air.
From these figures, you could go and calculate relative lethality data, which Hooligan has done in his laudable tables. But if you throw in the Aircraft damage model into the mix, you increase the number of variables considerably, many of which are unknown. I'd like to see someone construct a controlled experiment under those conditions.
If there would be a next step, for me it would involve confirming my results on a completely different setup, and testing the relative lethality against a non-hangar object (such as a city building), as I don't think the proportion guns/bombs holds steady between hangars and other ground objects.
Oh, and yeah, this test is really useful. Did you ever think about how much damage a b17 could actually do to an HQ building if, instead of bombing and leaving, it deacked, bombed and strafed? Or if some dweeb drove a cannon hawg up to the entrance?