Eventually on an update of vCenter/VCSA, the HA agent (FDM) gets updated as part of the activity. The FDM-package carries a zip file that will be copied by the FDM (upon starting the service) into a new offline depot of the update manager. If this action fails or happens too late, all hosts will try to gather the not (yet) existing zip and fail their side of the update. In larger environments this can quickly fill/flood the task-list. In the end if the zip-file does not become available (in time) the hosts will deactivate HA and be marked as failing in their respective HA group. Depending on spread of the issue this might lead to a high level of congestion on the remaining hosts with HA active in the respective cluster.
vSphere 7.0
vSphere 7.0.3
At first make sure the task-list becomes controllable again: either wait for the ongoing tasks to fail, which may take a long time as the hosts repeat the attempts a couple of times. Alternatively switch HA off for all clusters and reboot the vCenter in the usual way. The reboot of the vCenter flushes the task-list, disabling HA makes sure that the hosts do not immediately try to update again.
Then either of two actions may bring in the missing file:
In case of doubt ... just apply both.
Verify that the file is now available in the depot! In the vSphere 7 manual, this section describes the procedure of checking the depot's content:
Working with the vSphere Lifecycle Manager Depot
Once the file is found in place (i.e. in the offline-depot), it is time to continue.
In case it was necessary to stop HA, re-enable HA for the respective clusters.
The hosts should then update as expected. In case HA was not disabled, after restarting the FDM, they should begin to update. If HA was disabled, re-enabling HA will trigger the update.
After the updates ran through successfully, check the status of HA for all relevant hosts.