But, to be clear, you don't often create tables with dynamic names (though you will sometimes; for instance, to repeatedly load huge datasets, which comes up in time-series apps)... but you very often need to select which table to use at runtime based on user inputs. Again, in real-world apps.
For what it's worth, I believe the same quote_name() mistake pops up with column names too. But you probably don't dynamically select those based on inputs either.
ORMs are there to shield you from database details. I don't care whether the underlying database is a MySQL instance running on Amazon's cloud, an Oracle RAC down our datacenter or even a non-relational animal we use for very high volume operations. I don't care if the data is in tables or a tree. The kind of stunt you are trying to convince us is reasonable should never be done on top of the ORM, but under it, or, better, by its side. If you have a humongous dataset to load, don't do it through the ORM - do it directly on the datastore.
On a side note, I find your tendency to rely on name-calling and insinuations most disturbing and quite inappropriate for a place like HN.
The apps that I've seen do best against SQLI had an enforced policy that front-end devs couldn't write queries, that their access to the database was restricted through stored procedures, and that access was further restricted with database permissions. Stored procedures aren't a panacea either, but they're definitely more reliable than the Django ORM.
My problem with stored procedures is that when you go that way, your business logic will end-up inside the database or, worse, scattered through layers of application and database (ever tried to do version control on stored procedures?), something really bad when you need to know what is going on with the server or when you need to move to a different database.
If you check your ORM thoroughly (I never did that as thoroughly as I wish I had with Django) and you never, ever, write SQL in your business logic (something you really should go for) you will be in better shape than if you write SQL code that calls stored procedures. It's easier for the DBA to slip up in a your-business-specific way than for the ORM maintainer. Also, with the DBA, only a few people will have access to catch the mistakes whereas lots of people read (and abuse) ORM code regularly.
I think the most important point you mentioned is not the stored procedure, but the lock down. If the app isn't allowed to mangle the data it should not mangle, you are mostly good. The best part is that locking down the database can happen without touching the ORM and is more or less portable across different databases.