Roland's right, though: the error message looks identical to the Boost.Threads error that was fixed by this change. It would be nice if someone got to the root of the problem and figured out what change caused this error: it only appeared recently, at a time when no Boost.Threads changes had been made.
With all those replies, I was able to diagnose the problem. On a number of toolsets, <runtime-link>dynamic automatically sets <threading>multi. The program_options jamfiles had std::facet-support in requirements, which forces <runtime-link>static on intel. Apparently, the handling of std::facet-support is done after <threading>multi is added, which leads to inconsistency.
I'll raise the issue on jamboost@
BTW isn't the std::facet-support option irrelevant now? It only has any effect for Intel (last time I looked), and since Intel have long since fixed the bugs that caused us to add that option, can't we just get rid of it? John.