VARSOM-AM43 system on a chip defect after few weeks

Hello,

I have contacted Variscite on this matter but I don't get any usefull help.
I designed a new header board for the VARSOM-AM43 what is a system on a chip in a DDR socket.
The system runs very well for weeks in a normal environment inside an office at 25 degrees.
After a few weeks the SOM get's hotter and hotter until it will fail and shut itself down. (170 degrees)

I thought it would be some voltage spark from the power supply or something like that so I added a few TVS diodes to protect the input lines to the SOM. It looked like it was working but a month later a SOM was dead. Five SOMs are dead now.

I attached two files to make life easy:
- Evaluation board schematic of the VARSOM-AM43 (See page 9 for the power supply and input TVS diode)
- My own design for this VARSOM-AM43, it's the big 200 pin connector.

What I already did:
- Measurements with a scope on the supply rails, also under load and with no load, startup voltages (nothing weird found) voltage stays at 3,345V
- Added som TVS diodes to the supply rails of the VARSOM
- Contacted Variscite with this matter without any usefull reaction, they think it's software related....

Thanks for any reaction!