OK so I ran through the models to see how far I could "push" things.

The calculation factor (CF) will change a bit with the voltage, particularly at medium voltage. But I was just doing a preliminary numbers test so this only applies to 480 V. CF is inverted (higher values result in lower incident energies) so keep that in mind.

At 1244.6 and above the calculation factors end up at 1.21 for VCB, 1.24 for VCBB, and 1.14 for HCB. At 660.5 those are down to 1.12, 1.11, and 1.06. At 660.4 (a different formula applies) they are really close to the same numbers. Watching the equivalent lengths they are down to 26 mm across the board and continue to drop slowly to a minimum of 20 mm. At that point (508 mm) they stop decreasing and the CF's are at 1.0 across the board.

Using 18" distance and 0.1 seconds arcing time, varying bolted fault from 5 to 100 kA I get VCB from 0.51 to 6.06 cal/cm2, VCBB from 0.92 to 15.81, HCB from 1.72 to 22.88, VOA from 0.27 to 3.58, and HOA from 0.88 to 14.39. Running up to the maximum box size (1244.6 mm) these change to 0.4 to 5 for VCB, 0.7 to 12.7 for VCBB, 1.5 to 20 for HCB, and VOA and HOA are unchanged for obvious reasons (open air).

So it doesn't look like we ever reach a point where the box model equals the open air model which defies logic since as the box size increases eventually we would expect it to look exactly like an open air case other than maybe a factor of 2 higher due to being a hemispherical arrangement.

I took the liberty of reformatting your post for clarity with my response.

As the enclosure size increases, one would expect the results to approach the open air equivalent i.e. VCB=VOA and HCB=HOA as you suggest. However the missing factor is the back plane of the enclosure. Although the larger the enclosure size has less focusing effect on the incident energy (approaching open air) the back of the enclosure is still there which can reflect some energy towards the calorimeters resulting in VCB and HCB remaining a bit larger than VOA and HOA

Great observation!

Forgot about that. I was thinking about the other consideration that as distance increases the decrease in thermal radiation should eventually tend towards an inverse squared relationship when compared to itself, as the geometry tends more towards a point source. I was varying the wrong factor. However looking through everything the distance factor is still K12 which is the new "x". Distance is still basically a constant exponent that depends only on electrode configuration and not on distance.

Was hoping Wilkins ideas made it into the new equations because when you look at overhead switchgear or long distances with high power equipment and stick or other types of remote operation that are well over a 4-6 foot range, K12 should tend towards -2. At least the 1.3 factor that was present in IEEE 1584-2002 for larger gear has been replaced by something more realistic after considering enclosure size. His simplified model uses 2 circular discs and more of a thermal radiation theory to vary the energy with distance instead of simply applying polynomial curve fitting to it. Yes even theoretical models (Lee excepted) are realistically a mix of theory and curve fitting but they have a certain elegance to them. I'm sure though that this is getting into the pragmatists vs. theorists argument.