IEEE 1584 mentions that the electrode gaps were based on typical gaps found in equipment. I have seen references to that being based on a NEMA standard. I have searching for that NEMA document but do not know the number. Searches on NEMA website have not been fruitful.
Does anyone know what the document is that contains the spacings?

I don't believe the bus spacing are explicitly defined in the NEMA/ANSI standards. They require a certain voltage withstand rating and the spacing is left to the manufacturer. IIRC, the IEEE 1584 actually measured bus spacings of typical equipment of each type defined and used an average (or minimum?).

I believe that is correct. The various IEEE, NEMA, and UL equipment standards are performance specifications that, for example, may state that equipment has to pass a certain level of impulse response test. It is up to the manufacturer to establish gaps and clearances that are sufficient for the equipment to pass. There are various rules of thumb and field guidelines for clearances, some of which may be based on actual testing or manufacturer's internal design guidelines, but those gaps and clearances are not set forth in some definitive way in a standard.

Indeed, the numbers are published in the IEEE 1584 Table 2 - Classes of equipment and typical bus gaps.

There are no specific standards for minimum spacing. However that being said, the actual bus spacing from one manufacturer to another is remarkably similar, even at medium voltage levels.

For instance, most 480/600 V gear has a 25 mm (1") spacing, almost universally. There are similar dimensions for 5 kV, etc.

If you look at bus gap you will find two things stand out. An increase in bus gap necessarily (longer arcs) increases the incident energy. But (and this is where the arc flash models fall down) it also results in a less stable arc that often self-extinguishes. This is true because the air gap needs to maintain a temperature of at least 5,000 k or higher for the air to remain conductive. Otherwise as the current crosses through zero, the arc does not restrike.

Hi,

If the busses are coated with insulator like epoxy, how does that change the calculation? since arcing would be less likely to occur.?

IEEE 1584 does not consider arcing likelihood, only magnitude. IEEE 493 has limited data on arcing likelihood and breaks it down into phase-phase, and non-phase-phase. Most of the data is also very old. Do not assume epoxy bus is better. Arcs can travel across it, under it, and if it actually "stops" an arc, it can result in an increased incident energy from a plasma jet.