So to answer the questions:
Fascia collision:
I simulated the fascia on an imported Alembic Cache. Then I turned the bones and muscles into zBone-objects (Contact Stiffness: 1x106) and may fascia/fat mesh into a zTissue (Contact Stiffness: 1x106 / 107)
Because I was also thinking I might don't have enough thickness I made the fascia/fat mesh a bit thicker (thinner than the one of Mr. Ink but still pretty thick), but it still hat similar penetration issues.
For the sub steps I tried a lot of values which ranged from 4 - 32 but this had also rather minor effects.
The animation is not super fast. The arm is going from a straight pose to a pose where the arm is angled about 90° within 10 frames.
Yes I ramped on the values of pressure and surface tension in the way you showed in the tutorial video where you showed the new approach with the material layers.
Displaced vertices at the attachment locations:
Yes, I have a rig piping into the bones where I did the muscle simulation. I knew about the recommended workflow, but I kinda wanted to see why doing it in the rigged skeleton isn't recommended and which issues this is causing (and its a bit easier to see the results because I can tweak the testing animations faster).
Even though I am sure there are no cycles in the rig itself, but I guess you are rather referring to internal stuff Maya does, so I think this might most likely be the cause of this issue.
The vertices are not really dead, they follow the muscle tissue through the simulation like the do the right thing but with a fixed offset. (Parallel evaluation is turned on btw)
To answer the question about the tweaks:
I usually clean up the meshes to make sure they are properly zeroed out and don't have any history, but wasn't aware Maya actually saves the changes to the vertices in that kinda hidden place there. But I was testing around a lot to reproduce this issue, but wasn't able to do so. So I am pretty sure my meshes had not tweaks applied before turning them into tissues, but most definitely they had these offsets on the vertices that are saved in the CVs menu, because I was adjusting them to not intersect. But usually (I think) this shouldn't have much of an influence, because that's probably the case for a lot of other setups as well and I never saw someone reporting an issue like that. Also the purple triangulated mesh Ziva is producing still has the correct shape as you might can see on the pictures and it's only happening on the vertices that had an attachment, which is strange. I think there is always a tweak on the meshes as soon as you convert them into zTissues, but changing the envelope doesn't have any effect. When removing the zTissue though, changing the envelope of the corrupted meshes makes the vertices pop back into the right places (Which is the way how I fixed this issues the last days so far, but I can almost guarantee there was no tweak deformer applied before).
So to conclude:
I will set up the muscles in a separate file where they only simulate on the cache of the bones and see if the issue with the displaced vertices will stop to occur. (and will report back here if that's the case after some days)
In the case it keeps happening I will test with making sure there wont be any values in the CVs section with the method that James suggested and see if that will help.
For the collision issues with the fascia/fat simulation:
I will double check for each muscle if there might be a self intersection now, because it was simulated on the cached out muscles and bones and there might be a visually hidden intersection that was caused by the displaced attachment vertices, which then disables the collision for that specific muscle if I got that right?
Will let you know any results as soon as I have them.
Thanks a lot for the help!