If you still have the VM, you can try to back the VM up again (after making sure the last backup snapshot in the group failed verification, so the dirty bitmap is cleared). In case the original data is still there, the chunks will be uploaded again, which can "heal" snapshots which reference the same previously corrupt chunks.
If that does not work, then your snapshots are corrupt and you will have some data loss. You can nevertheless try to recover the raw image by running proxmox-backup-debug recover index <file> <chunks> --ignore-corrupt-chunks=true --ignore-missing-chunks=true --output-path=<path-to-target-image-file> on the PBS, <file> being the path to the index file (ending in .fidx in the respective snapshot folder) and <chunks> is the path to the chunk store base directory, so <datastore>/.chunks.
Above steps assume you are using a regular datastore, these will not work on an S3 backed datastore, there you will have to fetch the whole datastore contents from S3 by some external tooling first.
