Only in eight year chunks. By year seven there is more unalignment than there was in year one, but it goes back to normal on year eight. Same thing as with leap days, just a slightly bigger scale.
In fact, with current rules, [the shift in the regular Gregorian calendar becomes quite big when considering 100-year and 400-year cycles](File:Gregoriancalendarleap_solstice.svg). In theory, a leap week calendar with new and updated rules could have a very comparable if not a smaller average deviation from the true solar date, though I haven’t ran the precise calculations
Wouldn’t be many users, but would having a separate version on F-Droid be a way to be able to reach some of them?