Query fuzzers are a powerful tool for correctness, they cover the state space slowly though. We should investigate (among other things):
1. coverage guided fuzzing (https://github.com/google/fuzztest)
2. a more methodical crawl of the query state space via a DFS or BFS
3. speed up the running of the query fuzzers to cover more queries in less time
4. increase the "validity" rate of queries generated
5. cross-platform fuzzing: PM-3099