Post Go back to editing

How Can AD7415 device transients think it is a wrong I2C address?

I have an embedded application using AD7415-0 part.   I have 4 instances (in perhaps 2000 devices) where temp sensor stopped responding it I2C commands.   Probing on the I2C command from the master, it all looks good.   On 3 of the 4 devices, power cycling recovered the circuit.  On the 4th device, the I2C "address" for the device seems to have changed.   I'd like information at to how the AS pad works to "define" the I2C address for the device.   Is it polled only at power up to check its state (VDD, GND, or floating), or  is the "state" of the AS pin actively checked, either all the time or (dynamically) at the beginning of every I2C communication from the master.  My device is currently behaving as though AS pad is tied high instead of to GND.   I am wondering what will happen when I power cycle this device.   Do I have an internal flop that has undergone a SEU or gotten in latchup?   I have used this part in previous embedded application and have not seen this.