One more thing: I do not think it's appropriate to allow this in GET STACKED DIAGNOSTICS. That's about reporting the place where an error occurred, not the current location. Eventually it might be interesting to retrieve the OID of the function that contained the error, but that would be a pretty complicated patch and I am not sure it's worth it. In the meantime I think we should just forbid it.
If we do that, then the confusion you were concerned about upthread goes away and we could shorten the keyword back down to "pg_routine_oid", which seems like a good thing for our carpal tunnels.