Two out of those three are a real possibility though.

Now they are

That still sounds like really, really, really bad design. 40C is not a hurricane, flood, or volcano. It's a mildly hot day. Engineering fail.

As I understand you are engineering for a range of temperatures with problems to be found outside either end of the range (though I imagine it’s also possible to engineer for wider or narrower ranges to some extent, with appropriate cost implications).

Coping with higher maximum temperatures means more cracking and damage at very low temperatures, so it is not necessarily a case of just building in more buffer at the top end but a trade off between the cost and frequency of events like this and the cost of dealing with the effects of winter.

Given that this had literally never happened before and the frequency of similarly hot events was also lower in the past, it’s perhaps unsurprising that the trade-off landed where it did, even if it may well be in for a rethink now.