I have probed the Eeprom and at power up there is activity on Chip select, but clk and data both remain high - not due to being shorted to 3V3 but because of the 10k pull-ups. This is on the pins of the FT232HQ and the Eeprom.
It's tantilisingly close...
As to how I broke it:
I was using the analog inputs set to 100mV/div, differentially measuring two dc voltage outputs that were 0.25V - 1V.
The AD2 GND was also connected to the GND of the test circuit.
The circuit generating these outputs was powered from 18V bench supply - with a fair amount of transient current ability and a floating output, so GND was connecting thru the AD2.
I accidentally shorted something to something - I think I left it powered and put my grounded soldering iron on the 18V. Effectively I forced the 18V of the bench supply to be GND and the GND to try and be -18V.
The result was as above.
I'm guessing that at 100mV/div there is still attenuation in the analog input circuit, so I don't think the current flowed there, so it must have been between the AD2 input GND and the soldering iron GND, thru the powered and grounded USB hub.
Cheers
Alex