Design a Test Matrix for Feature Flag Combinations Without Testing Every Permutation

I am shipping a change that touches code gated by several feature flags. Help me decide which flag combinations deserve tests.

Flags involved and what each controls: [list each flag, default value, and a one-line description]
Code areas or modules that read these flags: [list]
Existing tests and which flag states they run under: [describe or paste names]
Known interactions or dependencies between flags: [describe, or say unknown]

Produce:
1. A table of flag interactions: for each pair, say whether the flags are independent, one depends on the other, or the interaction is unknown and why.
2. A minimal set of combinations to test, with the reason each one is included. Always include the current production state, the intended post-rollout state, and the all-off state used for rollback.
3. Combinations that are impossible or invalid in practice, and a suggestion for how the code should reject them.
4. For each chosen combination, two or three concrete assertions about behavior, not implementation.
5. Anything that cannot be covered by automated tests and should be checked manually or via a staged rollout.

Prefer pairwise coverage over the full cross product when flags are independent, and say explicitly when you are relying on that assumption.

Sign in to upvote this prompt.