Random csum errors

Dear devs,

since 26.07. scrub keeps reporting csum errors with random files.
I replaced these files from backups. Then deleted the snapshots that still contained the
the corrupt files. Snapshot with corrupt files I have determined with md5sum, here I get an input/output error.
Following new scrub, still finds new csum errors that did not exist before.

Beginning with Kernel 5.10.52, current 5.10.55
btrfs-progs 5.13

Disk layout with problems:

mdadm raid10 4xhdd => bcache => luks
mdadm raid6  4xhdd => bcache => luks

Already replaced 2 old hdds with high Raw_Read_Error_Rate values.

Aug 02 15:43:18 server kernel: BTRFS info (device dm-0): scrub: started on devid 1
Aug 02 15:46:06 server kernel: BTRFS warning (device dm-0): checksum error at logical 462380818432 on dev /dev/mapper/root, physical 31640150016, root 29539, inode 27412268, offset 131072, length 4096, links 1 (path: docker-volumes/mayan-edms/media/document_cache/804391c5-e3fe-4941-96dc-ecc0a1d5d8c9-23-1815-92bcac02c4a72586e21044c0b244b052f5747c7d2c25e6086ca89ca64098e3f3)
Aug 02 15:46:06 server kernel: BTRFS error (device dm-0): bdev /dev/mapper/root errs: wr 0, rd 0, flush 0, corrupt 414, gen 0
Aug 02 15:46:06 server kernel: BTRFS error (device dm-0): unable to fixup (regular) error at logical 462380818432 on dev /dev/mapper/root
Aug 02 15:47:25 server kernel: BTRFS info (device dm-0): scrub: finished on devid 1 with status: 0