So as a Rule of Thumb: We should not be relying on the ability to run any "all of" query, or really any query that we didn't specifically model for, and/or that doesn't otherwise have a reasonably bounded result set cardinality. This is basically a scatter-gather and full merge over all the nodes in the cluster. As the size of a dataset grows, these get very expensive (prohibitively so) and time consuming to run. It's also problematic because the set of [wiki_id,page_id] tuples here isn't deterministic, and we should assume it has changed by the time the query completes; I don't think this is a good strategy for this.

All of that said, the table is currently small enough to reasonably do a COPY from the cql shell, and I have attached the output of that below. Let me know if this will work.

Boldly closing this; Please reopen if there continues to be work needed here.