Is it possible that over time the heat may have expanded the material that the screws are set in? I recall when new homes were being built in the 80's/90's, drywall screws were popping through the mud affixing the sheets due to the green lumber drying over time. It's that that I envision the screw holes becoming larger due to heat, thus allowing the screws to slowly working their way out due to the constant, albeit slight, vibration.
Duly hope the above helps, guys. At least something to consider.
It's possible. When I removed the heat block to take a look underneath I snugged it down, ran it for 24 hours, then snugged again in a criss-cross pattern while it was running and warm. Just checked now, torque is the same so it didn't change. They do have additional screws holding the heat plate to the board, so the support is not all around the chips. Which is good.
When I was working on boosting people's jallies I would snug down the heat sink to the plate to finger-tight torque, let it run for a day, then snug it down to the final amount. Either the thermal compound spreads a bit or the threads expand a little. You can't crank down hard at first, because that will crack the chips. Just finger tight is what's needed, like a watch.
This is not totally uncommon, when I swapped an engine head on my old 944S Porsche I had to torque it down to the first spec, run it for a bit, then re-torque to the final values. After that it never changes.
Good analogy! Perhaps, the same principle is in place here considering the amount of heat produced.