I have been following this discussing for sometime now, as I am also facing this issue.
We have one scale out repository with two extents, coming from Windows 2016 servers with dedup enabled. All MS/Veeam patches applied, even the recently released Update 2 for Veeam 9.5.
The scale out repository is configured to use per-VM backup files. Windows Deduplication configued
There are two backup jobs configured in the environment, and at least once a fortnight there will a couple of VMs failing with "storage metadata corrupted" errors. The only way around it is to delete the affected VM from configuration/disk and start a new chain. But sometimes all VMs in the job will fail, and as such I have to start the chain from scratch. Very annoying and disappointing.