[nightpool]

Right, this is exactly what GP is talking about. When you prioritize "doing predictable sprints" and a stable point velocity over releasing features to production, you reduce your team's throughput and productivity. When you prioritize shipping features to production over "predictable sprints", then things can be successful. You're proving GPs point, not contradicting it.

[robertlagrant]

You're contrasting two different types of thing. If part of something being "done" is deploying it to production, then it will be useful to have the "deploy to production" bit of it managed by a sprint process that tries to account for all the work needed to do something.