Dwy, Thank you for the suggestions.
Moving a promoted solid slab: if I use a GC point as origin to the CS generated as origin or centroid of the slab, then the promoted solid will move as I move the GC point around without disappearing. (In any case, this points at some issues with CSs not generated by Cartesian coordinates and their "freedom" to be moved.)
As you wrote, "baking" a GC solid could be done by exporting as element into another DGN file (at this point all "GC" flavor would be washed off and only simply attributed DGN geometry would remain). When attaching the reference to a movable CS back into the GC model, then I can move the attachment around (and rotate, scale, etc, depending on how I control the CS). If I base my range points on that CS, then consistently "re-awakening" the solid to a GC solid is possible through the feature from element in attachment technique.
If I wanted to edit the solid in MS mode, most likely the exported file should be copied, and the solid be modified in the copied file in order to avoid over-writing modifications with further exports. Those modifications would be one-offs, though and any other changes upstream of that disruption in the pipeline would have to be propagated manually or are simply "lost". The attachment back into GC would reference the copied file with the modified solid.
Referencing of GC-enabled geometry into GC is another story. At this point one would have to use file inclusion which of course is pretty much like copying all the transactions from the included file into the active GC model.
Volker