LT4356 fried by minor overvoltage


we have an issue with a LT4356 getting fried during in-house lab testing. Its voltage clamp is set to about 39 V, and during static overvoltage conditions (tested by providing a continuous 45 V to the input), the IC malfunctions in a way that it stops driving the GATE above the OUT voltage. It is not a temporary condition, since after power cycle the condition stays the same for normal voltages (35 V and below). 

See schematic:

For testing purposes, TVS diode D40 was removed to prevent clamping the input voltage. Output current was minimal, cca 20 mA, during the whole test. The input voltage was increased from 35 V up to 45 V, when it stopped working. Gate voltage on Q4 was cca 12 V above the OUT voltage during the test, until it stopped working at 45 V input. Note that the output current wasn't changed at all during the whole test.

With the fried board, I tried the following: decouple LT4356 gate driver from transistors, and provide a fixed 35 V on its OUT pin (which then propagates to Vcc and SNS). The gate voltage still remains almost the same as OUT, doesn't rise above it. Is it possible that the internal schottky diode that protects the GATE pin got fused? What could cause this?