Thanks Cliff. What happens is that either they get an error uploading the package OR the system just sits there for a long while and gives up. After the unsuccessful attempt the system appears to be unstable for a while, reporting very high CPU usage. I tried importing their package into a HALO HCI VM and it was unsuccessful as well, so it doesn't look like the problem is in their cluster.
It's very likely they exported things that made no sense in the destination system. Looking at some JSON files I see references to the QA cluster name, for instance.
To make it more interesting, they unwisely got rid of the QA system before they imported the package into the production system so there is no way to redo the export,
Thanks anyway. Any ideas are more than welcome.