The real cost of a slow deploy

When deploying is hard, teams batch changes. Batched changes are harder to review, harder to test and harder to roll back — which makes deploying harder still. The loop tightens in the wrong direction.

The second-order cost is worse: engineers stop proposing small improvements, because the overhead of shipping one exceeds its value. The system ossifies, not because anyone decided it should, but because the pipeline made change expensive.

Fixing this is unglamorous infrastructure work — reproducible builds, real staging, automated tests worth trusting, one-command rollback. It reads like a cost line. It behaves like a capability.