Queries potentially read many nodes, for example because they don't use an index or because they use join conditions that return many nodes.
Such cases are sometimes not detected early because the default storage engine (the Tar storage) is fast. However it is a problem for slower storage engine (that stores data remotely).
To detect such cases early on, we could add a limit on how many nodes a query may read. It might make sense to use multiple distinct limits: one for how many nodes were read but skipped, and another limit for how many nodes were read, and a third for how many nodes are kept in memory (for example for "order by" queries).