Right now, when performing a query over a JS object, the entire object is recursively copied. For simple queries over large objects, this can get a little silly. For example, running the query count @ over a large array constructs a brand new array from scratch to query over, as executed in the input garden wall.
This is obviously not ideal. We should somehow be operating the MistQL semantics over the array that exists. It's certain that for some operations, we absolutely need to build a new array (e.g. @ | map @ + 1), but for simple queries, that shouldn't be the case.