Several Canadian provinces have vast forest management areas with hugely intricate rights (e.g. Alberta have dozens of first nations, dozens of oil companies and dozens of forestry companies all with different rights to the same overlapping areas), and many of them have huge datasets that needs to be coordinated. E.g. years back when we started doing this app, I visited with one forest management company that were doing logging in an area the size of Ireland.

Most of the polygons will be smaller, but there's any number of objects mapping out boundaries of massive areas to a ridiculous level of detail. More often the datasets will map out large number of moderate size (tens to thousands of points) polygons. Anything from buildings to traditional hunting trails, to oil and gas pipelines or mineral deposits.

Many of the huge polygons could be simplified just fine without any noticeable loss of precision, but that's not always a choice anyone is willing to take responsibility for doing, and so the system needs to work on whatever overly detailed data is passed in - customer knows best and all that - (but often the right answer for indexing this kind of data is just that: Simplify the outlines however drastically works for your application, with the most extreme being simply the bounding box; intersect that, discard all the "misses", and do the expensive full intersect on what's left).