>On a server with multiple VMs on a big consumer NVMe How did you avoid the crippling performance penalty with storing VM disks on btrfs? The usual workaround

>On a server with multiple VMs on a big consumer NVMe How did you avoid the crippling performance penalty with storing VM disks on btrfs? The usual workaround suggested to btrfs users is to disable copy-on-write for VM images, but doing so also turns off checksumming thus disabling the very features that make btrfs, well, btrfs.