That means,assuming the DPAK regulator had thermal shutdown, it could not switch off the regulator sufficiently to prevent solder melt. So it is wrong to use the PCB copper area as heatsink. Maybe the caps failed due to warm PCB?
Then why are more manufacturers switch to power packages that use the PCB as heatsink?
In fact right now I am looking at an advertisement of D class amp chip with 6W output max @85% efficiency, implying self dissipation of ~1W. And this device is packaged in a 5mm x 5mm x 0.8mm QFN pack!