No, this isn’t about crashes — but about getting timing drift down as low as possible by stabilizing the on-board crystal’s temperature.
You know, I have noticed that clock drift can be caused by overheating. So this makes sense.
For many years, some folks have been plotting the estimated freqency adjustment value (ordinarily written to the ntp.drift file) and also the temperature. There is generally a clear correlation.
Not only can one generally see how the temperature affects the frequency adjustment, with enough observation/data one can take the frequency adjustment and derive the expected temperature.
In an office setting, folks have long noticed how the frequency offset and temperature change, as buildings may disable their heating/cooling from (for example) 6pm and 7am on workdays, and also “over the weekend”.