VMware came back and told me that there was an issue because of a vpxa.cfg file on the ESXi hosts, but they don't have an idea of the cause. They did say that 'too many delta disks or cloned disks depending on other disks which adversely affects the vpxa agent' , which is interesting because there was a single snapshot on each of these machines which is the one Veeam was trying to create.
They wanted me to edit the vpxa file on each of the hosts and change 128 to 256 and restart vpxa then enable sfcdb temporarily for a core dump.
I'm not sure I want to proceed with this change given that their reason for doing it doesn't seem to be correct
I also have been told to change the VddkPreReadBufferSize on my case with the wording that
"it doesn't appear that the root cause of the issue has been found - however, it does appear to be related to ESXi hosts that are added via vCenter: the error itself seems to be produced due to an issue with the datastore or connection between the host and the datastore"
I'm somewhat surprised at that statement, since we had no issue on v7 or v8 whatsoever for nearly two years with no changes to our environment.
I'll be adjusting the VddkPrereadBufferSize value today but if Picaroons response is correct, I doubt we will see much difference