Comment # 3 on bug 957379 from Bruce Rogers

(In reply to Alexander Graf from comment #2)
> The make command seems to have spilled its address space with python calls
> and clones before the execve() to run make inside a subdirectory.
> 
> I don't think we really have to investigate this much further. We can simply
> do all the python calls explicitly before the qom test suite by running
> "make tests/qom-test" before the "make check-report.html" call.
> 
> Verifying whether it reproducibly fixes the build. If it does, I'll commit
> the change.

I tried this as well (semi-similar to the approach I stumbled upon of just
doing
the make check-report.html one more time if the first fails, which I saw
consistent success with), and this succeeds for me consistently, so I think you
are definitely on the right track here as a solution.