https://gcc.gnu.org/bugzilla/show_bug.cgi?id=124200
--- Comment #2 from Thomas Berger <loki at loki dot codes> --- I did my homework today and deep dived into the current name-lookup logic, and the handling of BINDING_VECTORS there, and i agree, every other approach, besides complete loading of all decls, would introduce changes to CMI or other core parts. As i can't think of any other reason to visit every BINDING_VECTOR in a namespace besides `members_of`, touching the CMI is surely overkill. I'll try to get something working.
