Double-squashed migration is omitted from forward migration plan after migrating backward

[Jacob Walls]

Start with this models.py and generate three simple migrations:

from django.db import models

class A(models.Model):
    foo = models.BooleanField()

./manage.py makemigrations 

from django.db import models

class A(models.Model):
    foo = models.BooleanField(default=True)

./manage.py makemigrations 

from django.db import models

class A(models.Model):
    foo = models.BooleanField(default=False)

./manage.py makemigrations 

Migrate.
Squash, migrate, double-squash, migrate.
Migrate back to zero.

./manage.py migrate
./manage.py squashmigrations new 0001 0002
./manage.py migrate
./manage.py squashmigrations new 0001_initial_squashed 0003
./manage.py migrate
./manage.py migrate new zero

Now notice the forward plan is missing a migration:

./manage.py migrate
Operations to perform:
  Apply all migrations: new
Running migrations:
  Applying new.0003_alter_a_foo... OK

And reverse migrate fails (failure would have been evident in the last step if it hadn't been an SQL noop):

./manage.py migrate new zero
Operations to perform:
  Unapply all migrations: new
Running migrations:
  Rendering model states... DONE
  Unapplying new.0001_initial_squashed_0002_alter_a_foo_squashed_0003_alter_a_foo...Traceback (most recent call last):
  File "/Users/jwalls/django/django/db/backends/utils.py", line 103, in _execute
    return self.cursor.execute(sql)
           ~~~~~~~~~~~~~~~~~~~^^^^^
psycopg2.errors.UndefinedTable: table "new_a" does not exist

[Jacob Walls]

I assume that the first squashed migration is not being unrecorded along with the second squashed migration when migrating to zero.

[Sarah Boyce]

Thank you!
Replicated, regression in 64b1ac7292c72d3551b2ad70b2a78c8fe4af3249 (refs #24529).