Ok, as discussed with Emmanuel Bernard, there will be no duplicates in the key array passed to getTuples(). This means we can abort the inner loop for assigning result rows to requested keys as soon as the first (and only) matching key has been found. That'll improve perf of this a bit.