The result most surprising to me is that laptops are between 25% and 60% less likely than desktop machines to crash from a hardware fault during the first 30 days worth of measurements.

The much larger weight and volume of desktops would seem to make them easier to cool.