Nice & Thanks.
Maybe good idea to pin this, or commit to CE
I think that several issues here are caused by corrupted data/fs.
Is the emmc also ejected before powerdown by CE/BL? This is a managed NAND, which may also have background maintenance running, which does not like brownout.
EDIT: both linux Kernel drivers support Power off notification, just checked. Hopefully the host notifies the mmc driver…