New numerical experiments with Sherards device. The porosity of fine particles is smaller now, that result that it is more rigid.
The mechanism seems to agree well with Sherards clarification at the beginning, but radial shift of particles from the hole looks more likely to be diagonal and depends on whether there are cracks inside the finer particles. If the width of the fine particle layer would be bigger, I think the agreement would be little better as upper part of finer particles will stay rigid.