Steve,
The term "cross-certification" comes from X.509 and has a
syntactic defintion provided there. Your neologism, "semantic
cross-certification" is, IMHO, a misleading choice of terms that is
easily confused with the term defined in international standards and
in ongoing ANSI (X9F1) work. I suggest you develop another term to
describe the latter concept.
Your own characterization of the unanswered, though not
necessarily unanswerable, issues assoiated with multi-root systems
seems to be largely in agreement with my observations. My suggestion
is that these issues are worth exploring, and when suitable answers
are in hand, then we can properly compare the current, well-defined
system codified in 1422, with alternatives that are equally
well-defined. In the meantime, it is potentially misleading to users
to promulgate a system that is a hybrid with unstated semantics [I
might say the system was neither fish nor fowl, but then I'd be
tempted to say it was closer to foul ;-)].
Steve