Environment Strategy
The deliberate choice of how many pre-production environments exist, what each is for, and how faithfully each mirrors production.
Most organisations accumulate environments rather than design them, and end up with four half-maintained tiers where nobody can say what dev proves that staging does not. The strategy question is what evidence each environment produces, and whether that evidence is worth its cost.
The expensive failure is parity drift. An environment with a fraction of the data, a single instance instead of a cluster, mocked third parties and different configuration cannot falsify the things that actually break in production: connection pool exhaustion, index selection on real cardinality, timeouts under real latency, quota limits on real volume. It produces confidence without evidence, which is worse than no environment.
The modern alternative is fewer long-lived environments plus ephemeral per-branch environments created from infrastructure code and destroyed on merge, with production risk managed by progressive delivery rather than by another static tier.
The honest test: name the last production incident a given environment would have caught. If nobody can, that environment is a cost centre and a false signal.