Is it possible to write a test case for this scenario if we can force reproduce it?

Which part actually makes the difference? It looks like this re-enabled a commented out exception and adds an optional parameter to a function call. How does that resolve the issue, where do those changes track to and what to they cause to happen differently?

Which part actually makes the difference? It looks like this re-enabled a commented out exception and adds an optional parameter to a function call. How does that resolve the issue, where do those changes track to and what to they cause to happen differently?

Yes, that's the change that caused the fix. Raising that exception dooms the using connection while a transaction is aborted.

Ok. At least I'm not missing something then. Is there any clue why that line was commented out? It's a quite important line to accidentally remove.

I couldn't find anything; it was commented out since the very beginning (August 2018 as I found) before publishing on GitHub. 😕

cc @saurabh500