i wonder how, in the long run, we can begin to hit the cache for the "nearest" artifact for an incremental compile.