How do noUncheckedIndexedAccess, exactOptionalPropertyTypes and TS 5.6+ checks harden 2025 codebases?

Question presented to candidate: "strict is already on in our codebase. Someone proposes adding noUncheckedIndexedAccess and exactOptionalPropertyTypes, and upgrading TypeScript for the newer 5.6+ checks. What does each one catch that strict does not, and how would you roll them out?"

What a strong answer should cover:

• noUncheckedIndexedAccess (TS 4.1): index-signature and array element reads get | undefined, closing the hole where arr[5] or record[key] is typed as always present; .at(), Map.get() and for...of are already honest; neither flag is part of strict
• exactOptionalPropertyTypes (TS 4.4): prop?: T means "may be missing", not "may be undefined", so { prop: undefined } is rejected (TS2375 / TS2412); it matters because spreads, in and Object.keys treat a present undefined differently from a missing key
• TS 5.6+ checks that need no flag or come with strict: always-truthy / never-nullish conditions (TS2872, TS2869), strictBuiltinIteratorReturn, plus 5.7 never-initialized variables and 5.8 per-branch checks in return conditionals; and noUncheckedSideEffectImports (5.6, on by default in 6.0)
• Version context: TypeScript 5.9 tsc --init turns both flags on, and TypeScript 6.0 makes strict the default
• Rollout: enable per package, fix with narrowing patterns (check then use a local, Map, omit keys instead of writing undefined), avoid ! as a blanket fix

Clarifying questions expected:

• Which TypeScript version is the codebase on, and can it upgrade?
• How large is the codebase, and is there a monorepo where flags can be enabled package by package?

Code / implementation expected: Optional — an index read that crashes under strict alone, and a spread that overwrites a default with undefined.