The costmap and random planner test and the action_utils test are flaky. We should get them back up to >80%.

Sometime we see the waypoints task fail, but that's much less frequently.

• planner - costmap
• planner - random
• action utils. Not sure why this started behaving much better today, but I'll take it.