Looking cool, thanks for sharing!
I think you could bring down the resolution of the fascia a lot. From memory the lion was at ~60k and I think I could have gotten away with ~35-40k. Try to make sure your remeshing has an even distribution of points if you can, and even concentrate the density of points in areas where you really want to see valleys come and go, like the delts and upper back legs. Some good tools for remeshing/ retopologizing:
Mesh Mixer (outputs triangles only, but that's fine too)
In the area you circled, looking at your muscle model again, I think you're missing a muscle. Here's a photo I took from Animal Anatomy for Artists:
The one circled is pectoralis profundus.
One thing I've been experimenting with, which you could try is coupling a tissue with the fascia.
To do this you would do the following:
Duplicate the fascia twice.
Project one of the duplicates in a little (use the sculpt polygon tool and flood in the normal direction)
Make sure there's no self-intersections on the duplicate
Make sure there's no intersections between the projected version, and the other copy
Combine the two layers and bridge the border vertices
Make sure all the normals are pointing out
Convert this volume to a tissue
Set its material to be super soft. Something like 5x101
Set its density to be super light, so I used 10.6 (1/100th of default)
Turn collisions off on the tissue
Make sure the tet resolution is enough to support most of the curvature you want to represent.
Select fascia, then tissue and make a fixed attachment. You can make its stiffness something like 5x104
I've found that this stabilizes the fascia quite a bit and stops it bending on itself. I've also been able to reduce the substeps down quite a bit too. Worth experimenting with if you have the time.
Here's another variation you can try:
If you already have a fat volume (the area between the fascia and the skin) you can use that instead of following the steps above, but in this case you would turn collisions OFF on the fascia and ON on the fat. That way you don't need to run a fascia/fat sim separately.
Collisions should perform better because they are now "two way" meaning that every collision test is volume-volume -- rather than with cloth, which is surface-volume.
You may lose some detail in your fascia this way, but you lose detail in the final skin when running fat over a detailed cloth sim anyway. If that makes sense.
Let me know if any of that's useful!