in reply to
Re^2: Attempt to free unreferenced scalar
in thread Attempt to free unreferenced scalar
According to perldiag, this message is considered a warning.
You shouldn't be dismissing the severity. The code is accessing a freed memory!
My current theory is that the free() call is failing even when I don't see any output. Is that assumption reasonable?
Accessing freed memory can lead to a wide variety of behaviour. It's a waste of time looking at this when you already know what leads to this memory corruption (accessing a freed scalar).
So far, I have not successfully isolated the problem.
An alternate approach would be to find the commit that fixed the problem. This might give a hint how to work around the problem in the versions before it was fixed.
There are tools to bisect history to find what a problem started occurring, or in this case, stopped occurring. I don't know how easy it would be to automatically install modules as part of the process.