I agree, with you, it's kind of a catch 22 when dealing with energized work.
I forgot to explain that all this equipment has been there for 40 + years. Nothing is new, this is all existing equipment. If I was designing it properly, I would totally provide something with a maintenance switch to reduce the arc flash incident energy, but unfortunately I need to work with what I have.
Also, the panel that is in questions is on the secondary side of a 750 kVA transformer, that is why the incident energy is so high at this location and seems to be dictated more by available fault current instead of upstream overcurrent protection trip time. I think SKM assumes a worst case of 2 seconds to trip on the secondary side and does not include the primary over current protection trip time.
Anyways, I'm just trying to provide them with a scenario that would allow them to work on this equipment live (under 40 cal/cm^2) and reduce down time at the data center.
The standard procedure is to shut off the HV side disconnect of the transformer. Recognize that if it goes to 2 seconds that means in the event of an arcing fault it will not trip AT ALL. Conceptually the assumption is that the victim is able to evacuate. Make sure that is actually possible. IEEE 1584 is clear on the escape issue.
I have used three ways to get it to trip. First is mount an OCPD between the transformer and panel board. At the sizes you are talking about it will be class L fuses. Mount in a junction box NOT a disconnect. The only time these will trip is arcing faults in the panelboard or shorts in it, both very unlikely: I have mounted a large MCCB on the air termination cabinet too on dry transformers but using an external fuse block is cleaner.
Second is use a breaker on the primary side. With a programmable one switch to a tighter trip curve either from a maintenance switch or based on a timer or when current drops below rated after a close command. This avoids inrush while maintaining good secondary side protection.
Third option needs either a breaker or with a fused disconnect at 13.5 kv on the primary side uses a “smart” fuse sold by Schneider: We just need a shunt trip. Mount bushing CTs directly on the transformer secondary bushings. Mount and program a relay for instantaneous trip at or above arcing fault current. So we shunt trip the main based on a secondary side fault. You can take this to the extreme and change the panel to main lug only using the relay as a “virtual breaker” scheme.
Fourth is either manually turning on instantaneous or if possible making it switchable on the primary side breaker. Again this screws up coordination and is a manual function but it is usually “free”.
With most of these options 8-12 cal/cm2 is easily done for no or very little cost and a lot less hassle. You should realistically make that your mitigation goal. Multilayer arc flash suits have a number of severe limitations. I was the one that put in the public input deleting the 40 cal “limitation”. There is no longer a Code upper limit. There are 150 cal suits from Oberon. But realistically all plants should have a practical limit. It has been proven repeatedly that all can be built or modified so that all energized work involving an arc flash potential can be done at under 8-12 cal/cm2. A couple IEEE papers clearly document this. Single layer arc flash PPE (AR shirts and pants plus face shield and balaclava) can be worn indefinitely in hot locations (foundries and refineries make it every day work wear) so only the face shield and mask need to be added. The only time you need the 40 cal suit is using the table based approach.
I’m a service engineer. All I work on is old dilapidated equipment, AFTER the local guys have already had a crack at it. We often get involved in retrofits. Be aware that 70E comes with an informational note that essentially invalidates your entire arc flash study if equipment is not properly maintained. If you service breakers after 40 years for the first time I start right off asking what spares the customer has and we discuss how far we are going to go before we abort servicing and testing because there is going to be a lot of failures.