This is particularly useful for rasters as no distributable filters are applied based on the sfc keys exactly matching the data geometry (tiles). In general, the assumption is that if a query has a filter set on it (ie. SpatialQueryFilter, SpatialTemporalQueryFilter, CQLQueryFilter) applying that particular filter is more applicable, and covers at least the filtering of the individual dimensions of the SFC. However, in the case that no query filter is provided, but the multi-dimensional constraints are given, multi-dimensional range decomposition is not going to be as effective of a filter as filtering the values of each individual dimension for a key - so instead of just accepting extra false positives, we apply an additional NumericIndexStrategy filter.