Django group nearby rows based on datetime

Question

In a Django model that stores an event datetime, I would like to group rows nearby in datetime. To clarify, each object has a date_time field and I can find the gap between objects in date time rather easily with:

    # We want a minimum date_time as the default for prev_date_time
    # but because Session.date_time is a DateFimeField and TZ aware
    # Django requires teh default to be TZ aware. And datetime.min 
    # cannot be made TZ awre (it crashes), adding one day prevents 
    # that crash and works and serves our purposes.
    min_date_time = make_aware(datetime.min + timedelta(days=1), "UTC")
    sessions = Session.objects.all().annotate(
                prev_date_time=Window(
                    expression=Lead('date_time', default=min_date_time),
                    order_by=F('date_time').desc()
                ),
                dt_difference=F('date_time') - F('prev_date_time')
            )

This works brilliantly.

Now, I would like to group these sessions such that any session with a dt_difference of over 1 day marks a group boundary.

I imagine two new annotations could do the trick, but I am struggling to write them and would ideally like to do so without resorting to raw SQL.

1. A new annotation that is equal to the date_time of the session when dt_difference is greater than one day, null otherwise.
2. A new annotation that fills all these nulls with the first non-null value (in the direction away from the gap) with this date time.

Best illustrated by example. I extract some sample data from this table as follows and illustrate, in which I show the date_time and the prev_date_time and dt_difference as per the annotations above, and illustrate two new annotations, start_group and stat_dt.

id	date_time	prev_date_tim	dt_difference	start_group	start_dt
577	"2021-07-03 23:59:00+10"	"2021-07-03 22:30:00+10"	"01:29:00"	null	"2021-07-03 19:45:00+10"
574	"2021-07-03 22:30:00+10"	"2021-07-03 22:30:00+10"	"00:00:00"	null	"2021-07-03 19:45:00+10"
576	"2021-07-03 22:30:00+10"	"2021-07-03 21:00:00+10"	"01:30:00"	null	"2021-07-03 19:45:00+10"
575	"2021-07-03 21:00:00+10"	"2021-07-03 19:45:40+10"	"01:14:20"	null	"2021-07-03 19:45:00+10"
572	"2021-07-03 19:45:40+10"	"2021-07-03 19:45:00+10"	"00:00:40"	null	"2021-07-03 19:45:00+10"
573	"2021-07-03 19:45:00+10"	"2021-06-19 22:15:00+10"	"13 days 21:30:00"	"2021-07-03 19:45:00+10"	"2021-07-03 19:45:00+10"
580	"2021-06-19 22:15:00+10"	"2021-06-19 20:45:00+10"	"01:30:00"	null	"2021-06-19 19:15:00+10"
579	"2021-06-19 20:45:00+10"	"2021-06-19 19:15:00+10"	"01:30:00"	null	"2021-06-19 19:15:00+10"
578	"2021-06-19 19:15:00+10"	"2021-06-05 23:59:00+10"	"13 days 19:16:00"	"2021-06-19 19:15:00+10"	"2021-06-19 19:15:00+10"
571	"2021-06-05 23:59:00+10"	"2021-06-05 22:00:00+10"	"01:59:00"	null	"2021-06-05 19:00:00+10"
569	"2021-06-05 22:00:00+10"	"2021-06-05 21:30:00+10"	"00:30:00"	null	"2021-06-05 19:00:00+10"
570	"2021-06-05 21:30:00+10"	"2021-06-05 21:10:00+10"	"00:20:00"	null	"2021-06-05 19:00:00+10"
568	"2021-06-05 21:10:00+10"	"2021-06-05 19:00:00+10"	"02:10:00"	null	"2021-06-05 19:00:00+10"
567	"2021-06-05 19:00:00+10"	"2021-05-22 18:15:00+10"	"14 days 00:45:00"	"2021-06-05 19:00:00+10"	"2021-06-05 19:00:00+10"

Now start_group is not needed per se, I simply envisage it as an intermediary, conceivably easy to annotate. The goal is start_dt, which we can then GROUP_BY on to group them as desired, with nearby sessions with a day+ long gap either side.

I have a feeling this is best done and likely possible with annotations, so that a single query on the database can then return aggregate information on the grouped sessions.