I don't think we used a pull-up, the down-stream attachment of the USART1 is to RS232 level converter which just reflects states of those pins. The loader initially configures the RX pin as an input to a TIM, and uses that to measure the timing via 0x7F data pattern before switching to USART mode. I haven't reviewed the loader's configuration of the pins, although I'd expect it to pull-up or down the pins it is monitoring internally so as to avoid it mistriggering. I'll take a look at the ROM next week to double check.
The processor is perhaps overkill in the application, but the die geometry of the newer STM32 is better, allowing for smaller packaging I can manage in manufacturing, plus the CM4 allows for a full JTAG and SWV debug channel, and 32-bit TIMs. My production guys don't like the wafer scale pitches, and the pin escape blows up the PCB layer/complexity/cost.
Had to move to a L433 for another design because the pin usages was getting too restrictive and poor muxing options from ST meant I couldn't get 3 USARTs out of the smaller one. ST could probably get better up-take on low pin count devices if someone spent a bit more time thinking about what peripherals need escaping properly/completely.
