What do you want to do with the subseconds, or what do you expect from them?
The processor is in STOP mode most of the time and is woken up by interrupts generated by the GPIOs. I want to measure the time between interrupts with an accuracy of less than 1 second. There is no LPTIM available in this PN. The only available time base is RTC.
While the 'G0 does not appear to have the infamous "BYPSHAD=0 may not lock time/date properly" erratum, it's probably still more practical (read: faster = resulting in lower active-time power consumption) to use BYPSHAD=1 and the read-twice-and-repeat-if-not-match mechanism, see eg. EngyCZ's implementation here (in second part of that article).