Interesting problem, electricpete, I hope you will keep us informed. My first thoughts are that the analysis here ruling out mechanical causes are quite logical. Ruling out the trip relay as a cause was a good step too. My second thoughts however, concern the instaneous trip and inverse time...
At my power plant, a nuke, we have 600VAC three phase delta ungrounded supplying lots of equipment, and no arcing problems as you are asking about. No, you aren't lucky, you just don't have hard grounds. The arcing over the capacitance of the insulation just isn't going to happen at these low...
The mechanic that we are interviewing to maintain these units says we are overloading each unit by 20%. He claims that if the power factor is above the nameplate 0.8, you have to de-rate the output kW.
I would take a look at the vendor documents, and especially a capability curve if available...
