Simplify test suites without weakening coverage

[robinebers]

TL;DR

Audit all 139 Swift test/support files and simplify 53 test files without weakening meaningful regression coverage. Reduce repository test/support code from 27,885 to 25,825 lines (2,060 fewer; 7.4%).

What was happening

• Provider, layout, widget, pricing, cache, and notification suites repeated equivalent cases and assertions across multiple files.
• Large JSON and filesystem fixtures obscured the behavior under test, while catalog snapshots and private implementation checks added maintenance cost.
• Coverage was sometimes duplicated in provider-specific tests even though shared mappers already exercised the same observable behavior.

What this changes

• Parameterize equivalent auth, error, pricing, threshold, reset, accessibility, and rendering scenarios while preserving genuinely distinct failure modes.
• Minimize provider payloads, scanner fixtures, persistence setup, and notification scaffolding across 53 test files.
• Preserve account ownership, credential permissions, HTTP 401/403/5xx classification, cache freshness and concurrency, cleanup and cancellation, alias canonicalization, and scanner-to-widget integration.
• Keep production code and documentation unchanged.

Heads-up

• Three existing live/parity tests remain skipped unless their opt-in environment variables and credentials are available.

Tests

• Affected suites: 685 XCTest cases across 70 classes, 0 failures, 3 expected skips.
• swift test: 1,190 app XCTest cases, 4 CLI XCTest cases, and 3 Swift Testing cases; 0 failures, 3 expected skips.
• swift build: passed.
• git diff --check: passed.

---

Note

Low Risk
Tests-only refactor: no app behavior, auth, or data-handling changes. Residual risk is slightly less isolated cases if a merged loop misses a previously separate assertion.

Overview
Shrinks and consolidates XCTest suites (~2k fewer test lines) with no production-code changes. Equivalent auth, mapping, scanner, layout, cache, and formatter cases are folded into table-driven loops; shared helpers (configuredAuthStore, makePersistence, makeSync, limitResource) replace copy-pasted setup.

Fixtures and assertions are tightened (smaller JSON/CSV payloads, map equality instead of per-field checks). A few overlapping tests are merged (e.g. unknown vs synthetic Claude models, Copilot extra-usage/org-seat variants, Codex reset-credit fallbacks) while keeping distinct failure modes. Some concurrency fixtures use fewer files; pin-cap tests use two providers instead of four.

^{Reviewed by Cursor Bugbot for commit ecffbc6. Bugbot is set up for automated code reviews on this repo. Configure here.}