Artifact: Trunk-Based vs. Feature-Branch Decision

ex-07 · exercises co-01 · two teams, two branching models, two different right answers.

Team	Shape of the work	Branching model	Cadence/team-size property that drives it
A two-person team shipping to production daily	Small, frequent, low-blast-radius changes; both engineers see every change land within hours	Trunk-based development -- short-lived branches (hours, not days), merged straight to main	Daily release cadence makes a long-lived branch's merge cost (days of drift) larger than the coordination cost of two people integrating constantly; a two-person team can also review every change fast enough to keep trunk green.
A five-person team on a quarterly release train	Large, coordinated feature sets that must all land together for one dated release	Feature branches (a release-scoped integration branch per quarter)	A quarterly cadence means release isolation genuinely matters -- an in-progress feature must not leak into a release train it is not ready for, and five people's parallel work needs a place to integrate before the release gate without touching main.

Verify: each pick cites the specific cadence/team-size property (daily release cadence for team A; quarterly release isolation for team B) that drove it -- not a blanket "trunk-based is always better" or "GitFlow is always safer" -- satisfying co-01's rule that the right branching model depends on team size and release cadence, not fashion.

Key takeaway: trunk-based development wins when integration frequency matters more than release isolation; feature branches win when the reverse is true -- and both of these teams are making the correct choice for their own shape of work.

Why It Matters: a team that copies trunk-based development from a blog post without a daily (or faster) release cadence to match it gets all of the coordination discipline trunk-based dev demands and none of the payoff -- while a team that copies GitFlow without needing release isolation pays merge-conflict cost for a safety property it never actually uses.