found that, and discussed it (with BB, in this forum) some years ago.

further investigations, conclusions and solutions can be found in my tutorial series on cloth room: www.book.artbeeweb.nl/?book=cloth-room. Part III (which is the "how to get your PhD in Poser Cloth Simulation", so do have some asperines at hand) especially chapters 3.3, 3.4 about.

For short: this crumbling is a sort of "mechanical resonance" in the simulation (math details in part V of the tutorial, it's completely), which becomes noticeably apparant in quad meshes, at low vertex density (#verts per cm2), at low friction (esp fold resistence). 

Hence, as the suggestions and findings above reveal: it helps to lose the quads and do triangles instead, it helps to increase vertex density by subsividing, and it helps to raise the resistance although Fold Resistance is more effective than Stretch (as the crumbling is a folding issue, not a stretching one). 
Triangles will make the cloth behave more like linen instead of rubber (eg woven instead of non-woven), while increasing vertex density or raising resistance will make the cloth behave thicker and more rigid.

have fun! 

- - - - - 

Usually I'm wrong. But to be effective and efficient, I don't need to be correct or accurate.

visit www.aRtBeeWeb.nl (works) or Missing Manuals (tutorials & reviews) - both need an update though