fix(catalog): load Alliance libraries with bounded paging

[zortos293]

Audit and implementation against the latest dev base, including 783db41f. The full evidence-backed comparison with OpenNOW-Mac is in docs/alliance-library-audit.md.

Library delivery and recovery

The core previously aggregated up to 2,000 games into one result while Qt rejects protocol lines over 1 MiB. Library now uses complete cursor pages capped at 100 games and 768 KiB, sharing the Store page-budget implementation. Oversized pages retry at the same cursor with a smaller count, without skipping rows.

Qt appends/deduplicates pages progressively, retains search on retry, cancels stale account/disconnect requests, and rejects cyclic or malformed pagination. Missing apps/items are errors rather than empty success. Desktop and console preserve partial results alongside a visible error and retry action.

Mac reference parity

• Honor explicit vpcId / vpc_id before the server-ID fallback. Route discovery through the configured proxy and separate metadata caches by route; subscription uses the same route.
• Preserve integer catalog IDs and server-defined ownership states, including selected variants.
• Refresh known expired ID tokens independently of access-token expiry, and share safe token selection across catalog, CloudMatch, linked accounts and storage. An expired ID token omitted by refresh is not retained.
• Keep existing provider endpoint restrictions, OAuth profiles, GraphQL identity headers and ownership filters. No guessed Alliance identity headers or public-catalog substitution.

Verification

• Full Rust core suite: 301 passed, 4 ignored; strict all-target Clippy and formatting passed.
• Full application/native build on Qt 6.8.3.
• 16/16 focused CTests, covering 24 QML paging checks, 78 orchestration checks, error surfaces, library routes, collections and Store paging/navigation.
• Localization validation passed for 1,483 QML source texts; only the English source was edited.
• Local HTTPS fixtures cover provider discovery, authorization, response mapping, paging budgets, malformed data, cancellation, proxy isolation and ID-token refresh.

The affected external tester's authenticated response is still unavailable. These are demonstrated client defects and reference-parity fixes, not proof that the tester's account now loads. The separate RTSPS Transport investigation in #883 remains outside this PR.

Visual acceptance: mocked partial-library failures, not a live account

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Advanced

Run ID: 26c8c033-540b-4fe9-b3df-7bac3d999dd8

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[capy-ai]

Capy found no issues.

View 2 other findings in Capy.