LTC2950I Unexplained behaviour at -40C.

I have a design that uses the LTC2950I for power on-off control. I am now testing the design at -40C and the unit is failing to power-up at this temperature. On inspection of the signals it seems the PB is being detected fine and the EN signal is asserted HIGH. However, after only about 50ms the enable signal drops back LOW. This does not provide enough time for the processor to de-assert the KILL signal so the unit can't start.

From what I can tell, the EN signal should remain high for at least 512ms under all conditions. So, I am wondering whether this could actually be the commercial variant LTC2950C that is only rated to 0C minimum. My supply chain has evidence that this was from a franchised source and was the industrial part at least on paper.

Does anyone have any experience with this? What else could explain this unexpected change of state?

My EN net has a 910k pull-up to supply around 8.5V and connected direct to PIC input.

Thanks,
B O