it's quick to get by PK and horribly slow to retrieve by date, but PKs increase monotonically with dates

crappy fake binary search.py

# there's probably some way to teach postgres to do this itself

import datetime

# convenience scribbles
dd = [x - datetime.timedelta(days=1) for x in [datetime.date(2021, y, 1) for y in [4,5,6,7,8,9]]]

def print_quick(*args):
    print(*args, end='', flush=True)

def find(date_value, date_field='date_created', model=Instance, pk=0, stride=10**6):
    # Pretty slapshod; if you include `pk`, make sure it's less than what
    # you're trying to find; if your `stride` overshoots the largest PK,
    # you'll be lost in space
    while True:
        i = model.objects.filter(pk=pk).only('pk', date_field)
        if not i:
            pk += 1
            print_quick('!')
            continue
        i = i[0]
        if getattr(i, date_field).date() > date_value:
            pk -= stride
            print_quick('>')
            stride //= 2
        else:
            pk += stride
            print_quick('.')
        if not stride:
            print()
            # This should be the first submission on the day after
            # `date_created`
            print(i.pk, getattr(i, date_field))
            break
    return
    while True:
        # Find the previous submission just to double-check
        # that it arrived on the previous day
        i = model.objects.filter(pk=pk).only('pk', date_field)
        if not i:
            pk -= 1
            print_quick('!')
            continue
        i = i[0]
        print()
        print(i.pk, getattr(i, date_field))
        break