Did you manage to figure something additional out? Do we know if this
also kills plain freebsd?
I'm afraid I'm stuck with this.
I narrowed it down to that particular test (or prerequisites of the
test) when executed by the GCC testsuite. A large chroot is needed,
having all the GCC build-deps; once GCC is finished building it's
fairly easy to trigger this on demand.
Even if it might trigger the bug on plain FreeBSD, I think all this is
too unwieldy to provide as a testcase yet (0.5GiB for the GNU/kFreeBSD
chroot and 4GiB for the GCC build tree).