Note that Linked Data caches work quite differently from the way as it is known from plain HTTP retrieval. An example how to cache Linked Data resources is implemented in the open-source software Apache Marmotta
There is no reason to license linked data access from commercial catalog providers. All the data should be free and open (another story is licensing software and paying for support). When using restrictive licensing on data, catalog linking will not work. This is the reason why many national libraries are providing authority files by using a CC0 waiver.
Transporting such files which are publicly available through the library community (and through the vendors' silos) will be easier than ever. First, with the internet, we have enormous network bandwidth capacities available for cultural data. Second, with Linked Data Fragments - see also
you can shovel the data you need very fast into your local system and prepare it for display or editing.
This is very different from traditional MARC data supply path. You are no longer tied to tedious file import/export or slow and unstable SPARQL endpoints. Linked Data fragments may be retrieved once per day, or once per hour - the target is the real time web, with the help of large aggregating linked data hubs, to distribute the authoritative information globally at the highest possible speed and reliability.