Sorry for not getting back to you earlier. Glad the screenshots gave you more insight into my problem, and thanks for your help!
Yeah, there were some changes to the UI in R25 (I think). But once you get accustomed to it, it’s okay. Icons are different, but IMO look simpler and neat, and you can still change the position of the tabs as you see fit. Their new cloth simulation system is probably one of the best updates.
I tried to increase the thickness of the stripes, but that didn’t really affect it. Unfortunately, changing the hierarchy didn’t help either. As you said, the deformation happening to the collider was probably too fast and big, so I used a delay effector for the animation and tried to slow it down a bit. My current settings are 100 steps per frame and max. solver iterations is set to 100 as well, error threshold 1%, collision margin 0.1 cm. These things helped, but 10% of the spheres still fell through. So my final changes to make it work were changing the friction from 70-80% to 10%, both on collider and the rigid body + bounce not above 5-8%. That helped especially when the spheres were set to Moving mesh. Convex hull worked even without the last changes of the friction and bounce.
Now 99% of the spheres are not falling through (using a convex hull).
Thanks a lot for the help and explanation about the limits of the dynamics.