Why Write Small CLs?
google.github.io
google.github.io
One thing I've noticed with small PRs is that the overall code quality stagnates. Reviewers focus on the small change, so they miss opportunities for abstraction and near duplicates proliferate.
Of course large changes have problems of their own, so I don't know what the solution is.
Mechanically, open CLs started retaining edit history, so they felt more like feature branches than just staging areas. I think Perforce style branches (rename my subtree using these rules) still existed but were discouraged and not well supported.