Current Lua access to data like index proofreading statistics invokes a dependency on every page, existing or not, in the index, because if any page is created or changes status, it can change the rendered data (note: the pages do not need to be actually accessed for per-index statistics, get just need the dependency so that the stats can be re-fetched en bloc from pr_index).

This is good for accurate rendering, but has very poor performance when data for many indexes (on the order of depending on 10000 total pages) are accessed. Sometimes, it may be a valid choice to trade off the accuracy for faster pages. So the Lua API could allow an "approximate" access mode that does not add the dependencies.