Greybeard
Indeed you are correct, avionics manufacturers employ an environmental stress testing in order to "weed out" weaker components and bring out other issues e.g. poor soldering. However issues can still appear post-test ater the unit has left the factory floor and is in service.
What kills electronics is thermal cycling - rapid heating up and cooling down. The most common way this can happen is the continuous cycling of power - i.e. switching on and off. Most failures occur at switch-on; at switch-on there are brief spikes of voltage and current which unavoidably over time damage the components, eventually causing failure. Obviously switching on-and-off also causes the components to heat up and then cool down again respectively; even in mil-grade components this causes damage over the long-term.
An interesting point to note is the decreasing availablility of mil-spec components; both suppliers and customers are growing increasingly wary of the high price (and, for suppliers, the low demand and hence small market) of mil-spec components, so increasingly avionics manufacturers are using mass-produced commercially-available components which can be found in anything from home computers to alarm clocks, and hence not spec'd to the more extreme temperature tolerances that mil-spec components have. Beancounters brilliance of course. Another reason why the free-market isn't always the answer to everything, despite what others think
With this in mind, it is easy to see why it is not a very good idea to continuously switch a radar on and off when cruising at FL300+ and the OAT is minus f-ing cold. Doing so on a regular basis over time will stress the radar components and will ensure an early visit to the workshop or a return to the factory. The radar installations I deal with are still in non-air-conditioned/nonpressurised environments (although typically processing equipment tends to be otherwise), although they usually have their own cooling apparatus for operation at higher ambient temperatures.
Smithy