It occurs to me that it may not be possible to achieve
the "single-SIDP criterion" directly in a case where
different, non-cooperating models are included in a new
model that encompasses them.  However, in this case,
the new encompassing model can have SIDPs whose
exhibited values are "derived" from the values
exhibited by the SIDPs defined in the included models,
in such a way that the derived SIDPs meet the
"single-SIDP criterion".  This suggests that we need an
additional step in the merging process defined by the
RM, in which these derived property values are
calculated.  It also suggests that the values from
which they were derived should be ignored by the
merging process thereafter.  Which, in turn, suggests
that the merging process must proceed hierarchically,
from the most deeply-included leaf models to the root
model that includes them all.  What do you think of
*that*?
If we do all that, we can make the "single-SIDP
criterion" a requirement for all model designs.