Anybody else seeing this odd behaviour around VM api backups in NBU7.7.2 it didn't seem to happen in the previous version(s).
Since ungrading from NBU 220.127.116.11 to 7.7.2 seeing a reasonable amount of VM consolidation errors ( VM 5.5) the bulk seem to be false errors that are real time wasters. The bulk seem to have no locks on then and can be consolidated from vCenter itself by logging in as the NBU authorisation user account. Those that refuse to be consolidated and report file locks they can be fixed by stopping any hung bpbkar processes on the media servers.
Failed backup reports generated by OpsCenter often list 4275 errors and these appear to be transient and fixed by automated reruns in the backup window. The other worry is that not all failed consolidation is reported. Backups often fail with error code of one. The machine is backed up OK but the parent/control job fails and reports failed consolidation. We pick these up by running a script against vCenter to report on machine that need consolidation.
Considering you went from NBU 18.104.22.168 (uses VDDK 5.5) to NBU 7.7.2 (uses VDDK 6.0) I'm curious on what exact level is your ESXi environment. As you stated "odd behaviour around VM api" there may lay a bug in code.
VDDK 6.0 release notes for you to look at https://www.vmware.com/support/developer/vddk/vddk-600-releasenotes.html