IO Error but badblocks and fsck are happy

Question

I have an ext4 partition with the following underlaying stack:

• sda1 and sdb1 are together in a RAID1, resulting in md0
• md0 is LUKS-encrypted, resulting in md0_crypt
• on top of md0_crypt is a single LVM volume mv0_vg_media, mounted under /home/media

When performing cp /home/media/hierarchy/photo.jpg /tmp I get an IO error (but just for like 20 files, out of several ten thousands).

Hoever, when I attempt to debug the problem:

• Both dmesg and the syslog stay clean when the IO Error occurs. (EDIT: clarification: This means that there is not any disk-related output logged in dmesg or syslog, even if the log is active with --follow option while the read errors occur.)
• Badblocks on sda and sdb do not reveal any errors
• fsck on /dev/mapper/md0_vg_media only outputs "could be narrower. IGNORED." warnings but no errors and the autocorrect option does not fix anything.

I'm puzzled. I could just delete those files and re-sync them, but that's a bad idea if I don't know what the problem actually is.

How can I further debug this?

EDIT:

From the comments and further research, I tried the following approaches (in vain) with the following results:

• mdadm --examine /dev/sda (same for sdb) returns: /dev/sda: MBR Magic : aa55 Partition[0] : 3907029167 sectors at 1 (type ee)
• mdadm --examine-badblocks /dev/sda (same for sdb) returns: mdadm: mbr metadata does not support badblocks
• cat /sys/block/md0/md/mismatch_cnt contains the value 0 (zero)
• echo 'check' > /sys/block/md0/md/sync_action performs the check but does not reveal any errors. Dmesg has two entries: [734796.807172] md: data-check of RAID array md0 and then immediately below: [754370.977181] md: md0: data-check done.