"No known data errors" despite CKSUM errors in a single disk setup

[smortex]

The host is a laptop with a single Samsung SSD 850 disk. This disk has a bunch of partitions, one of them being used for a ZFS pool. The output of zpool status is:

  pool: ztank
 state: ONLINE
status: One or more devices has experienced an unrecoverable error.  An
	attempt was made to correct the error.  Applications are unaffected.
action: Determine if the device needs to be replaced, and clear the errors
	using 'zpool clear' or replace the device with 'zpool replace'.
   see: https://openzfs.github.io/openzfs-docs/msg/ZFS-8000-9P
  scan: scrub repaired 0B in 00:03:04 with 0 errors on Sat Aug  9 12:27:05 2025
config:

	NAME        STATE     READ WRITE CKSUM
	ztank       ONLINE       0     0     0
	  sda4      ONLINE       0     0     1

errors: No known data errors

I am puzzled by the CKSUM error and the overall status "No known data errors": in my understanding, CKSUM increase when data is (successfully) read from the disk, but zfs detect that the checksum of the read data does not match the expected checksum for this data. When this happen and a system has multiple copies of the data, the data can be read from another device, and if the checksum is correct the now valid data is send to the first device to replace the invalid data and fix the issue.

But this system does not have redundancy, so what is described above and documented in the link provided with the error message in the above output cannot happen.

I did not find reference to something that would retry the same read from the same disk in hope this time the read data is valid (and I guess that would hit some cache at the hardware level that would make any different outcome improbable). I though that the error could have existed in a file that was removed since the checksum error happened, and so as the file with invalid content is now gone there are no data errors anymore… but this situation occur again and again, making this explanation weaker and weaker.

The disk has been scrubed and nothing changed (no more checksum errors, and still no known data errors).

What would explain this situation when a checksum was detected, could not be recovered, all files are readable and the pool still says there is no data errors?

[IvanVolosyuk]

Another possibility is that the error is in metadata and some parts of metadata stored with double and triple redundancy even on single disk configurations.