No that is correct. Users who run their jobs through Workflow send their resources to Workflow, who can then upload them on-demand to the Connect server. But you can very well upload the resources directly to the Connect server without going through Workflow first. So there can be a discrepancy between what’s stored on the Connect server and what’s on the Workflow server.
That’s because the Connect Server and the Workflow Server are not necessarily installed on the same PC. The servers are independent from each other. In fact, my own personal preference is to install those servers on separate machines to ensure they do not compete for the same hardware resources.
You could have several Workflow Servers sending jobs to a single Connect server, just like you could have a single Workflow Server sending jobs to multiple Connect Servers. That’s why the resources are stored in separate locations.