Excellent, that worked!
Just FYI, by changing the VM order in the backup job, the job would start VM1 on Host1, then error out on the next 3 VMs (from other hosts), then start VM2 on Host1 concurrently, then error out on the next 3 VMs (from other hosts), ...
So, I reconfigured each host for 1 concurrent backup, deleted the backup job, and recreated it with the default VM order (VM1 from host1, VM2 from host1, VM3 from host1, VM4 from host1, VM1 from host2, VM2 from host2, ...), and it worked as required. So, it's now backing up 4 VMs concurrently but 1 from each host. Throughput is much better than when I started above.
Sam Boutros, Senior Consultant, Software Logic, KOP, PA
Powershell: Learn it before it's an emergency