ZFS: `class=dio_verify_rd` events logged in syslog

I am using ZFS on an Ubuntu 24.04 host and have raw VM images stored on a dataset. QEMU/KVM uses them via

I've had this running like this for several years, and in April (~3 Months ago) I upgraded Ubuntu, and with that ZFS, which is now at this version:

modinfo zfs | grep version
version:        2.3.1-1ubuntu1
srcversion:     3CBA3F639287005F1F50BB3
vermagic:       6.14.0-24-generic SMP preempt mod_unload modversions

zfs --version
zfs-2.2.2-0ubuntu9.3
zfs-kmod-2.3.1-1ubuntu1

Why the mismatch between userland (2.2.2) and kernel module (2.3.1) versions? Direct i/o is enabled, if this is the correct query:

cat /sys/module/zfs/parameters/zfs_dio_enabled 2>/dev/null || echo "Parameter not found"
1

Since today I am getting tons of entries in syslog of the following type

2025-07-24T23:01:28.720398+02:00 host zed: eid=1131 class=dio_verify_rd pool='tank' size=131072 offset=623438361600 priority=0 err=0 flags=0x100280080 bookmark=3737:2:0:254570
2025-07-24T23:01:28.720526+02:00 host zed: eid=1132 class=dio_verify_rd pool='tank' size=131072 offset=623445984256 priority=0 err=0 flags=0x100280080 bookmark=3737:2:0:238679
2025-07-24T23:01:30.287882+02:00 host zed: eid=1133 class=dio_verify_rd pool='tank' size=131072 offset=2604696819712 priority=0 err=0 flags=0x100280080 bookmark=3737:2:0:187028

zpool status shows no issues:

zpool status -v tank
  pool: tank
 state: ONLINE
  scan: scrub repaired 0B in 00:30:08 with 0 errors on Tue Jul 15 04:30:09 2025
config:

        NAME                                          STATE     READ WRITE CKSUM
        tank                                          ONLINE       0     0     0
          mirror-0                                    ONLINE       0     0     0
            nvme-WD_BLACK_SN850X_4000GB_2xxxxxxxxxxx  ONLINE       0     0     0
            nvme-WD_BLACK_SN850X_4000GB_2xxxxxxxxxxx  ONLINE       0     0     0

errors: No known data errors

The host hosts 7 VMs, 5 are Ubuntu 22.04/24.04, one is Windows 7 and the one causing the logged events is Windows 10. Only the Windows 10 VM shows this symptom. The VM booted just fine and was usable without noticeable issues (but I don't use it much). I have now changed the QEMU/KVM disk config to cache=writeback and the log entries ceased to appear. Are there known issues with this new ZFS feature? Should I disable it globally at ZFS level? If so, how would I do that?