LM5116: Possible part latch up occurring at customer facility

We are using the LM5116 in a buck topology to provide a 700mA current source for an an inductive load.  Our VIN is a 70V DC rail and our output voltage is around 30volts with a nominal load.  We have a high side current sense resistor with circuitry to produce the single ended FB signal.  We have built this design and it has been working fine for several years, but recently we have been getting back failed units that have been isolated to the LM5116 (5 to date).  Nothing fails but this part.  The symptom of part failure is that our external VCCX (set at 8.8V)  is being loaded down by excessive current draw from the part.  The current required to bring up VCCX is several hundred mA compared to 6mA typical.  We have noticed that the LO pin to PGND resistance is much lower (around 100 ohms vs 400kohms) than normal.  The high VCCX current only occurs when the part is enabled and is normal when disabled.  Our AGND and PGND plane connections are around a inch from the part which is not optimal.  Our TI rep suspects latch up possibly due to our grounds not being connected together right at the part.  I have been looking at the part to determine what pins may be sensitive to latch up.  The part seems very robust with respect to latchup for AGND and PGND.  I notice that the FB pin can latch up the part and load down the VCCX rail, but this current is there all the time (both on and off) and the LO to PGND resistance is normal.  Question:  What pins if driven below PGND can cause the part to latch up in this manner?  I have tried the LO and VOUT pins in the lab and have not been able to latch up the part with these.  Also, should we be concerned about the plane connection 1 inch away from the part in terms of this failure mechanism described?  We do not have a snubber on the FET outputs like shown in datasheet.  Is this something that can cause negative transients that may cause part to latch up?  Thank you for any insight you may have on this issue.