i've been facing this issue a few times when backing up a pool of VMs which belong to a cluster or are Active Directory Domain controller and so on:
i understood it was 'safer' to save these machines not at the same time to avoid that the (brief) stun time at snapshot removal happens at the same time for all the machines, causing a downtime of the service or other problems.
I've workarounded this saving the VMs in two different jobs, or forcing a job to use the same proxy to reduce the chance of a simultaneous snapshot commit.
I'd love to have a switch "process VMs sequentially" in job settings
In case all VMs have a single vmdk, a possible workaround will be to limit the amount of slots on the proxy that is used for the backup job to 1 slot thus forcing the job to process VMs one by one.
