Is there a possibility to have that fixed rather than using the workaround?
If you can prove the regression (in what v11 patch release - i.e., before 13 - did your code work, show your work please) we'd be at least willing to consider something that could be back-patched to restore the old behavior (which shouldn't have changed in a minor release anyway). But at the moment we haven't observed this (I personally haven't tested though, the bug reporter is expected to help out here).
The problem is with ORM frameworks where there is no possibility to use this clause OVERRIDING SYSTEM VALUE ...
As there is an ORM involved, the claim of this being a regression in PostgreSQL is indeed worthy of doubt, as it may be the ORM that is at fault here.