Eli Zaretskii <eliz@gnu.org> writes:

> Once again, all the crashes are inside memory-allocation functions,
> which suggests some kind of memory corruption.  Did someone try to run
> this scenario under valgrind?

I've tried it now, log attached (minus what I believe are some false
positives that printed during startup).  This is against latest master
[1: 5352bda4ee].