The heat is not unpredictable, but it's not simple either. You'd have to use differential equations to calculate heat distribution and dissipation. It may not be as simple as you described especially when you want to predict the actual temperature (not just claim it'll get hotter).
You make a fair point, the details are not simple, but to me the basic concepts are mostly. Electronic component heat follows a conduction path from hot to cold to a heat sink surface area, and then convection to ambient air.
Predicting a temperature is best done by software like Solidworks or Ansys. No one I know hates themselves enough to manually crank out differential equations on a complex system like a circuit board and its heat paths.
My point, maybe badly expressed, was really that eventually a camera is going to get too hot with the sun burning down on it unless it has faster heat dumping ability than the source, and I find that impossible with these tiny bodies, so I stand by asserting any MILC will eventually get generically 'hotter' and eventually hotter is 'too hot'.