We have rare but serious failures of the LNK304 used in a high voltage buck converter. Input is a 450V DC bus that is generated by a boost PFC. Failure mode is a melt-down but not a hard short as seen by heating of adjacent traces for long enough for PCB traces to burn the FR-4. The circuit schematic is attached. I think that the use of a ceramic input capacitor to replace an electrolytic has increased the frequency of failures. Any ideas as to the cause? We have had some mechanical damage at this location on the PCBA by our CM mis-handling the PCBA, which caused us to change to a Boardflex 1KV rated input bypass capacitor.