#68957 is a performance problem caused by
O(n) lookup for associated items by name. It seems that searching through all associated items for one with a particular name is a common pattern. The obvious solution is to use a better data structure, but the result of the
associated_items query needs to be interned. Do we currently intern hashmaps or btrees? If we don't and it's hard to do, we could sort the list of associated items by name before interning.
cc @Matthew Jasper