https://gcc.gnu.org/bugzilla/show_bug.cgi?id=123870
--- Comment #10 from Jeffrey A. Law <law at gcc dot gnu.org> --- I know I've fixed problems in that space as well (validate_change_or_fail) related to thead-vector. So Robin's note applies there too -- ie, there's a good chance that's already fixed and would need to be identified/backported as well.
