ASR Azure-to-Azure on Trusted Launch Ubuntu 24.04: involflt "Root disk not found in driver context for TVM" keeps logging

Environment

• Azure-to-Azure replication, East US to West US 2
• Source VM: Trusted Launch (Secure Boot + vTPM), encryption at host enabled, Ubuntu 24.04 LTS
• Kernel 6.11.0-1018-azure (listed as supported in the A2A support matrix), pinned with apt holds
• Mobility service 9.67.7789.1, extension updates set to manual (no newer version is offered)
• Single 30 GB OS disk (StandardSSD), no data disks

Symptom

The kernel log repeatedly shows:

involflt[process_iobarrier_tag_volume_ioctl:3404 (ERR)]: Root disk not found in driver context for TVM and tag commit is pending
involflt[revoke_tags_timeout:3172 (ERR)]: Starting revoke tags timeout procedure

During the worst period, vacp -systemlevel -cc -parallel exited with status 4 after about 6 minutes, and vacp.log showed:

CheckDiskRestoreOrSwap Error: disk id to lun map is empty

Timeline

• The errors started the day replication was enabled, at about 230-360 a day since then.
• Replication later went Critical (errors 153006 and 153007, RPO about 4.5 h) because no crash-consistent points were created.
• We restarted the agent service, killed a hung vacp, fully rebooted the VM, and let ASR resync.
• Since the reboot, replication is Healthy: crash-consistent points every 5 minutes, RPO under 1 minute. However, the involflt errors came back within 1 minute of boot and are still being logged.

Ruled out: disk space (41% used), outbound connectivity to the cache storage account and Entra ID, data change rate (VM is close to idle), kernel version (certified, no drift).

Questions

1. What makes involflt lose the root-disk context on a Trusted Launch VM, and is this a known issue with 9.67 on Ubuntu 24.04?
2. While crash-consistent points are being created, are these errors a risk to recovery-point consistency or failover?
3. Is there a fix (a newer Mobility agent, a driver setting) or a recommended workaround?

Happy to share more logs if needed.