STM32 gotchas
128. In 'F2, subseconds don't work in RTC, nor smooth calibration

The STM32F205/207 models were concieved as improved versions of the then high-end STM32F105/107 models. The 'F2 received improvementes in many peripherals (e.g. the GPIO busmatrix), more FLASH and RAM (including battery-backup SRAM) and higher clock frequency.

Soon after STM32F205/207, the STM32F405/407 appeared, boasting an improved Cortex-M4 core with single-precision floating-point unit, and even more increased clock frequency. Contrary to the 'F4 line, the 'F2 line has never seen any more newer/better models; but given that the 'F2 is in almost all aspects (except the core and clock frequency) identical to 'F2, the 'F2 is considered a "somewhat worse 'F405/'F407" since then.

However, those who "downgrade" to the 'F205/207 from 'F405/407 for cost, availability or any other reason, may stumble upon one of the rare differences in peripherals of those two models: the lack of subseconds in the 'F2 RTC. Subseconds (and the related registers for alarm subseconds etc.) are simply not implemented in the 'F2 RTC.

Another relatively often used feature also not implemented in the 'F2 RTC is the smooth calibration. Btw. this may be the reason why coarse calibration is implemented in 'F4 - for upward compatibility.

There are more "defficiencies" in the 'F2 RTC as compared to 'F4 RTC, in the less-used functionalities: missing 1Hz calibration output, synchronization feature, shadow registers bypass, filtering/sampling on tamper input etc. Similar "reduced capability" RTC module is in the Cat.1 'L1 family, but that's less likely to cause surprise when migrating, as 'L1 "subfamilies" have higher mutual differences in several other peripherals/aspects than 'F4 vs. 'F2.

AN4759 contains an excellent overview of capabilities of RTC modules implemented in various STM32 families (except the simplistic rev.1 RTC in the ancient 'F1 family).