I agree. It is difficult to cause a power MOSFET to go into thermal runaway, under normal conditions, that is. But operating a TO220 MOSFET at near a watt where it is not a major source of resistance in the circuit and providing no cooling at all is one way this can be achieved.
Looking at the normalized Rds(on) vs. temperature curve for an IRL540 and taking it piece by piece shows us that at 850mw, the temperature of the 540 will want to rise by 52 degrees in free air at an ambient of 25 degrees. At 77 degrees (25 + 52) the Rds will rise by a factor 30%, increasing the dissipation to 1.1w, causing the temperature to rise to 93, causing the dissipation to go to 1.4W, causing the temp to rise to 111, causing....The curve actually just gets steeper and steeper.
And all that assumes free air with a fixed ambient temperature of 25. Sealed in a plastic box, this will happen much quicker as the ambient temperature will rise, being heated by the MOSFET.
The self-stabilizing effect becomes prominent when the MOSFET itself is a major contributor to the circuit's overall resistance. Then a 1.5 factor increase in Rds will reduce the overall current dramatically and actually lower the dissipation of the device.
The fact that the load in this case is the major contributor to the circuits resistance makes the matter worse since the IRL540 going from .08 ohms to .12 ohms will have little effect on the drain-source current, but a large effect on the power dissipated by the device.
Actually, the load here is quite likely going to be a constant current load too.