It's not so much the change in size but how much constraint there is to that change in size. It does not take much change in size to ramp up the stress if that change is constrained. For example, if we use 30E6 psi for young's modulus, 17.3E-6 in/in/°C for the thermal expansion coefficient of stainless steel and a 1 foot x 1 foot door panel, you get some pretty significant stresses if you are constraining that miniscule expansion. A thin enough panel will bow/buckle to accomadate that expansion. Over time, flexing of the panel within a constraining frame will experience local yielding that can accumulate.