Our bucketing strategy should not leave us with one bucket having 16 tests while two others have zero tests.
Furthermore, we may not need to use hashing at all. We already calculate the set of tests available for bucketing (via stuff like suite inclusion/exclusion), so we should just have each bucket take 1/b tests (where b = total number of buckets).
See Issue #165 for background.
Our bucketing strategy should not leave us with one bucket having 16 tests while two others have zero tests.
Furthermore, we may not need to use hashing at all. We already calculate the set of tests available for bucketing (via stuff like suite inclusion/exclusion), so we should just have each bucket take 1/b tests (where b = total number of buckets).