I propose a simple split of systests into about 5 large buckets. That
way, when one hits a failure, one doesn't have to rerun the entire
gigantic pile.

Reply via email to