I wonder why they make the xtals flat around +20C, when wearing the watch warms it up well above that - around +30C.
Actually they are flat around 25C. It is a laboratory standard temperature, but usefully it's quite close to ideal for a wristwatch because most watches are only worn for maybe half to two-thirds of a day, and quite often unworn for several days. The on-wrist temperature varies a lot, depending on how tight the strap is, etc. But 25C might be considered a roughly "average" temperature.
Even in a clock radio it will be annoying.
Really? My "radio-controlled" clocks and watches synchronise every 24 hours.
I believe wrist watches use micro RTC calibration (the 32F4 etc support this but I never used it) and the pricey ones perform it over temperature.
Yep: active temperature compensation is used on the most accurate watches. In terms of setting the mean rate, in many watches the crystal is laser-trimmed during manufacture - the classic Swatch was the first volume watch to use a laser-trimmed crystal. Otherwise an inhibition circuit is used. The crystal runs a shade fast, and the electronics drop one or more cycles at regular intervals. The degree of inhibition is set up by burning one or more narrow traces on the the circuitry.