Hi Sol,
thanks for the suggestions but I think I found my problem.
I haven't noticed that the mfs-nio-driver also does some initializiation of the sdhc and sdcard.
My fault was to add a sdhc-pe-component to init it, but this causes some errors. It is curious that the system worked 90%.
If in the future the MFS will also be a pe-component, this combination should be checked/excluded.
I don't know if a check is now possible when pe and non-pe is mixed.
Regards
PS: I read some posts about people who changed clock and bit-width of the sdhc without any changes in performance. Maybe they also did the same like I and configured the sdhc-driver instead of the nio-sdhc-driver. ;-)