[ZingerLittleBee]

Summary

This audit found tests exercising helpers and mutation paths that production no longer calls, alongside genuine behavior bugs. Remove only paths with no production callers and a traced replacement or current domain contract. Move meaningful assertions to active components, API handlers, transactional alert admission, and real client requests.

• Fix alert polling that kept postponing repeat notifications, preserve SSH and scan event windows after expiration, and reject invalid journal byte arrays.
• Deduplicate network history by timestamp, retain chronological ordering, and prevent stale iOS history requests from overwriting a newly selected range or clearing its loading state.
• Align widget SDK uptime, alert, and monitor hooks with server DTOs, disable uptime requests without a selected server, and add SDK type checks and tests to CI.
• Retire orphan chart, markdown, color, parser, cache, identity-merge, and iOS paths after checking callers and replacement history. Preserve enrollment, authentication, APNs, and wire compatibility coverage. Remove unused direct web dependencies without unrelated lockfile upgrades.

The five commits separate SDK, web, iOS, agent, and server changes. Regression tests were observed failing before the corresponding behavior fixes; deleted legacy tests were replaced or retained on active interfaces where their assertions remain meaningful.

Validation

Passed locally:

• Rust workspace library, binary, and example tests: 2,626 passed. All 46 non-mobile server integration targets: 852 passed.
• Workspace Clippy across tests, examples, benches, and all features with warnings denied; Linux agent cross-target compile check.
• iOS 26.5 full suite: 473 passed, zero failures or skips, including fresh server-to-relay trace fixtures.
• Widget SDK: 56 tests and type check. Relay: 74 tests, type check, and Wrangler dry run. Tooling: 29 tests. Renewal Python harness: 18 tests.
• Web production build and type check, documentation contracts and types, installer and changelog checks, navigation and integration-target guards.
• API regeneration produces no OpenAPI or generated TypeScript diff. Changed JS/TS/JSON files pass targeted lint; changed Rust sections pass formatting review; diff whitespace check passes.

Final JS checks used the repository-pinned Bun 1.3.4 and Node 24.13.1.

Remaining blockers and evidence limits

The complete local gates are not all green. The failing tests remain intact:

• Web full suite: 869 passed / 3 timeouts, all in unchanged renewal editor tests. The same renewal suite subsequently passed 30/30 in isolation with unchanged timeout budgets.
• Mobile push integration full suite: 85 passed / 5 failed on cooldown timing, reconnect, and message deadlines. All five failed cases passed in serial focused reruns. The full-suite failure remains unresolved.
• iOS 27 full suite: 472 passed / 1 timeout awaiting a notification delegate callback (74.776 seconds against a 60-second budget). The same case passed on iOS 26.5; the runtime difference does not establish its root cause.
• Whole-repository lint encounters 35 existing errors in ignored local artifacts. Crate-wide formatting checks encounter existing untouched formatting debt. Neither is hidden by broad cleanup or relaxed checks.
• Existing public SDK useHistory and useTraffic contract defects still need a separate API and range-semantics decision; their supported exports are retained.
• Linux sensor capture with privileges, physical-device APNs delivery, and the renewal UI scheme were not verified. Linux evidence is compile-only; simulator and static boundary checks do not establish device delivery or signed entitlements.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
serverbee-docs	Ready	Preview	Oct 17, 2026 5:40pm UTC