There was a resolution after the following study:
to consider the intrinsic size of regions to be that of its generated content only. The thread suggest to instead go with option 1. in the study (i.e., base the intrinsic size on a one time measurement of all the content, i.e., spanning all regions).
There is some agreement that it is better to have this solution than no shrink-wrapping behavior at all.
Chaning the component to be "Regions".